research paper on poverty and crime

The Official Journal of the Pan-Pacific Association of Input-Output Studies (PAPAIOS)

• Open access
• Published: 05 June 2020

Dynamic linkages between poverty, inequality, crime, and social expenditures in a panel of 16 countries: two-step GMM estimates

• Muhammad Khalid Anser 1 ,
• Zahid Yousaf 2 ,
• Abdelmohsen A. Nassani 3 ,
• Saad M. Alotaibi 3 ,
• Ahmad Kabbani 4 &
• Khalid Zaman 5  

Journal of Economic Structures volume  9 , Article number:  43 ( 2020 ) Cite this article

153k Accesses

46 Citations

269 Altmetric

Metrics details

The study examines the relationship between growth–inequality–poverty (GIP) triangle and crime rate under the premises of inverted U-shaped Kuznets curve and pro-poor growth scenario in a panel of 16 diversified countries, over a period of 1990–2014. The study employed panel Generalized Method of Moments (GMM) estimator for robust inferences. The results show that there is (i) no/flat relationship between per capita income and crime rate; (ii) U-shaped relationship between poverty headcount and per capita income and (iii) inverted U-shaped relationship between income inequality and economic growth in a panel of selected countries. Income inequality and unemployment rate increases crime rate while trade openness supports to decrease crime rate. Crime rate substantially increases income inequality while health expenditures decrease poverty headcount ratio. Per capita income is influenced by high poverty incidence, whereas health expenditures and trade factor both amplify per capita income across countries. The results of pro-poor growth analysis show that though the crime rate decreases in the years 2000–2004 and 2010–2014, while the growth phase was anti-poor due to unequal distribution of income. Pro-poor education and health trickle down to the lower income strata group for the years 2010–2014, as education and health reforms considerably reduce crime rate during the time period.

1 Introduction

The study evaluated different United Nation sustainable development goals (SDGs), i.e., goals 1 and 2 (poverty reduction and hunger), goals 3 and 4 (promotion of health and education), goal 10 (reduced inequalities), and goal 16 (reduction of violence, peace and justice) to access pro-poor growth and crime reduction in a panel of 16 heterogeneous countries. The discussion of crime rate in pro-poor growth (PPG) agenda remains absent in the economic development literature, though Bourguignon ( 2000 ) stressed to reduce crime and violence by judicious income distribution; however, a very limited literature is available to emphasize the need of social safety nets for vulnerable peoples that should be included in the pro-growth policy agenda for broad-based economic growth. Kelly ( 2000 ) investigated the relationship between income inequality (INC_INEQ) and urban crime, and found that INC_INEQ is the strong predictor to influence violent crime rather than property crime, while poverty (POV) and economic growth (EG) significantly affect on property crime rather than violent crime. The policies should be developed for equitable income and sound EG for reducing POV and crime across the globe. Drèze and Khera ( 2000 ) examined the inter-district variations of intentional homicides rate (IHR) in India for the period of 1981 and found that there is no significant relationship between urbanization/poverty and murder rates, while literacy rate has a strong impact to reduce criminal violence in India. The results further indicate the lower murder rate in those districts where female to male ratio is comparatively high. The study emphasized the need to reduce crime, violence and homicides by significant growth policies for sustained EG in India. Neumayer ( 2003 ) investigated the long-run relationship between political governance, economic policies and IHR using the panel of 117 selected countries for the period of 1980–1997 and concluded that IHR can be reduce by good economic and political policies. The results specified that higher income level, good civic sense, sound EG, and higher level of democracy all are connected with the lower homicides rate in a panel of countries. The study emphasized the need to improve governance indicators in order to lowering the IHR across the globe. Jacobs and Richardson ( 2008 ) examined the interrelationship between INC_INEQ and IHR in a panel of 14 developed democracies nation and found that intentional homicides is the mounting concerns in those nations where the inequitable income distribution exists, while results further provoke the presence of young males associated with the higher murder rates in a region. The policies should be formulated caution with care while devising for judicious income distribution with demographic variables in the pro-growth agenda. Sachsida et al. ( 2010 ) found inertial effect on criminality and confirmed the positive relationship between INC_INEQ, urbanization and IHR. The study emphasized the importance of public security spending to reduce IHR in Brazil. Pridemore ( 2011 ) re-assessed the relationship between POV, INC_INEQ and IHR in a cross-national panel of US states and found POV-homicides’ linkages rather than inequality-homicides’ association. The study argued that there is substantially desire to re-assess the inequality-homicides’ linkages as it might be the misspecification of the model. Ulriksen ( 2012 ) examined the relationship between PPG, POV reduction and social security policies in the context of Botswana and found that broad-based social security policies have a significant impact to reduce POV, thus there is a strong need to include social security protections in the pro-poor growth (PPG) agenda for lowering the POV rates across the globe. Ouimet ( 2012 ) investigated the impact of socio-economic factors on IHR in a panel of 165 countries for the period 2010 and found that GIP triangle are strongly connected with the IHR for all countries, while for sub-samples, the results only support the inequality-homicides association rather than POV and EG induced IHR. The results highlighted the importance of GIP triangle to reduce IHR in a panel of selected countries.

Liu et al. ( 2013 ) investigated the relationship between national scale indicators of socio-economic and demographic factors and crime rates in 32 Mexican states and found that EG, wages and unemployment negatively affect crime rates, while increase federal police force that is helpful to reduce crime rates; however, on the other way around, higher public security expenditures are linked with the higher crime rates in Mexican states. Chu and Tusalem ( 2013 ) investigated the role of state to reduce IHR in a panel of 183 nations and found that political instability increases IHR, while anocracies is the strong predictor to influence IHR in a panel of countries. The study concluded that IHR increases in those countries where there is high level of political instability and death penalty, while the amalgamation of democratic and autocratic features lead to increased IHR. The policies should be drawn to strengthen political governance across the globe. Adeleye ( 2014 ) evaluated the different determinants of INC_INEQ in a large panel of 137 countries using the time series data from 2000 to 2012 and found that per capita income (PCI), secondary education, rule of law index and unemployment rate are the strong predictors for INC_INEQ and IHR, while INC_INEQ considerably affected IHR rate in a region. Dalberis ( 2015 ) investigated the relationship between INC_INEQ, POV and crime rates in Latin American countries and found that INC_INEQ has no significant association with the crime rate in Colombia, Brazil, Uruguay and Salvador, while poverty is the strong predictor to influence crime in Brazil, Uruguay and Salvador. The results highlighted the need for pro-poorness of growth reforms that would be helpful to lowering the crime rates in Latin American countries. Harris and Vermaak ( 2015 ) considered the relationship between expenditures’ inequality and IHRe across 52 districts of South Africa and found that while keeping other district features constant, inequality does appear as a strong dominant player to induce IHR. The rational income distribution along with broad-based EG may play a vital role to reduce IHR in South Africa. Stamatel ( 2016 ) investigated the relationship between democratic cultural values and IHR in a panel of 33 democratic countries for the period 2010 and found that democratic cultural values have a positive and negative impact of IHR in the presence of strong democratic institutions and practices. Ahmed et al. ( 2016 ) identified the different predictors of economic and natural resources in the context of Iran using the time series data from 1965–2011 and found that labor productivity, exports, capital stock and natural resources are the main predictors of EG, which altogether are important for sustained long-term growth of the country. Enamorado et al. ( 2016 ) interlinked crime rates with higher INC_INEQ using a 20-year dataset of more than 2000 Mexican municipalities and confirmed the causal relationships between the two stated factors. The results confined that drug-related crime rates largely increase up to 36% if there is one-point increment in the INC_INEQ during the specified time period. The study concludes with the fact that drug-related violent crime rates are more severe due to high proliferation of large dispersion in the labor market in terms of negative job opportunities in illegal sector. Thus, the sound policies are imperative to seize drug trafficking organizations by force for pro-equality growth. Ling et al. ( 2017 ) analyzed the role of trade openness in Malaysian life expectancy using the data from 1960 to 2014. The results show that continued EG and trade openness substantially increase life expectancy during the study time period. Further, the results established the feedback relationship between income and life expectancy in a country. The study concludes that life expectancy may increase through imported healthcare goods, which improves the quality of life of the people, thus trade liberalization policies are imperative for healthy and wealthy wellbeing.

Zaman ( 2018 ) extensively surveyed the large weighted sample of intellectuals about crime–poverty nexus and explored the number of socio-economic factors that concerned with high crime rate and POV incidence in Pakistan, including INC_INEQ, injustice, unemployment, low spending on education and health, price hikes, etc. There is a high need to increase social spending on education and health infrastructure in order to combat POV and crime rates in a given country. Imran et al. ( 2018 ) considered a time series data of US for a period of 1965–2016 and concluded that incidence of POV increases the intensity of property crime in a given country, while other controlling factors including country’s PCI and unemployment rate are not significantly associated with property crime in a country. The study concludes that property crime should be restricted by strong legislative and regulatory measures, judicious income distribution, and increasing minimum wage rate, which altogether would be helpful for the poor to reap economic benefits from PPG reforms in a country. Zaman et al. ( 2019 ) evaluated the role of education in crime reduction in a panel of 21 countries for a period of 1990–2015 and found a parabola relationship between PCI and crime rates in the presence of quality education and equitable justice across countries. The study further confirmed few other causal conceptions among the variables for making sound policy implications in the context of criminal justice. Piatkowska ( 2020 ) examined the social cost of POV in terms of increasing suicides rates, crime rates, and total violent rates in the United States and across 15 European nations during the period of 1993–2000. The results show that suicides–crime–violent rates are substantially increasing due to increase in relative POV and infant mortality rates across countries. The study argued that relative POV is the strong predictor to increase social cost of nation that needs efficient economic policies to reduce crime rates. Mukherjee ( 2019 ) discussed the role of social sustainability in achieving economic sustainability by reducing different forms of violent/crime rates through state intervention in the context of Indian economy by utilizing the data for a period of 2005–2016. The results further highlighted the need of socio-economic infrastructure development that would be helpful to provide safety nets to the poor in order to reduce crime rates in a country. Duque and McKnight ( 2019 ) presented the channel through which crime rates and legal system provide a pathway to increase INC_INEQ and POV across countries. The study further discussed and highlighted the socio-economic vulnerability that escalates through unequal distribution of income and high POV incidence, which need effective legal system to reduce crime rates. Khan et al. ( 2019a ) surveyed the Bolivian economy to assess pro-poor environmental reforms that could improve the quality of life of the poor through judicious income distribution and sustainable environmental reforms. The results conclude that services’ sector and healthcare infrastructure would be helpful to reduce POV rate and achieve PPG process at country wide. Zaman et al. ( 2020 ) surveyed the large panel of countries (i.e., 124 countries) for a period of 2010–2013 to analyze the role of INC_INEQ and EG on POV incidence across countries. The results generally favor the strong linkages among the three stated factors to support GIP triangle, which forms PPG process. The study emphasized the need to adopt some re-corrective measures in order to provide social safety nets and income distribution in order to make a growth process more pro-poor. Kousar et al. ( 2019 ) confined its finding in favor of POV reduction through managing international remittances’ receipts and financial development that would be helpful to improve the mechanism of income distribution in a country like Pakistan. The study concluded that international remittances may play a vital role to reduce POV via the mediation of financial development in a country.

The real problem is how to make EG more equitable, which is helpful to reduce POV and crime rates, and make a growth more pro-poor. The SDGs largely provoked the need to sustained economic activities, which helpful to make growth policies more poor friendly. The previous studies are widely discussed crime rates and POV reduction (see Zaman 2018 ; Khan et al. 2015 ; Heinemann and Verner 2006 ; etc.); however, a very few studies interlinked POV–crime nexus under PPG and Kuznets curve (KC) hypothesis (see Saasa 2018 ; Berens and Gelepithis 2018 , etc.). Based on the interconnections between crime, POV, and PPG, the study formulated the following research questions, i.e.,

Does crime rate negatively influenced GIP triangle, which sabotages the process of PPG?

The recent study of Khan et al. ( 2019b ) provoked the need of PPG policies to ensure sustainability agenda by including socio-economic and environmental factors in policy formulation, which gives favor to the poor as compared to the non-poor. In the similar lines, the social spending on education and healthcare infrastructure, and reforms needed to reduce labor market uncertainty in the form of lessen unemployment rate is considered the viable option for crime and POV reduction across countries (Khan et al. 2017 ). Thus, the study evaluated the question, i.e.,

To what extent social spending on education, health, and labor market are helpful to reduce crime rate, poverty, and income inequality across countries?

This question would be equally benefited to the developmental economists and policy makers to devise a healthy and wealthy policy by increasing spending on social infrastructure for pro-equality growth (Wang 2017 ). The last question is based upon non-linear formulation of crime–POV nexus where it is evaluated as a second-order coefficient to check the parabola relationship between them, i.e.,

Does crime and poverty exhibit a parabola relationship between them?

The question is all about the second-order condition, which confirmed one out of three conditions, i.e., either it is accepted an inverted U-shaped or U-shaped or flat relationship between them. The second-order condition assessed the probability to reduce crime rates and incidence of POV in policy formulation.

In the light of SDGs, the study explored the impact of GIP triangle and crime rates on pro-growth and PPG policies, which is imperative for sustainable development across countries. The study added social expenditures in PPG dynamics to promote healthy and wealthy economic activities, which improves quality of life of the poor and helpful to reduce crime incidence across countries. The study is first in nature, as authors’ knowledge, which included GIP triangle and crime rate in PPG framework, while controlling different socio-economic factors, including education and health expenditures, unemployment rate, and trade openness. Further, an empirical contribution of the study is to include second-order coefficient of PCI for evaluating crime- and inequality-induced KC, while the study proceed to analyze forecast relationship between the crime and POV incidence over a next 10-year time period. Finally, the study estimated PPG index while including crime rate as a main predictor factor in GIP triangle for robust policy inferences. Thus, these objectives are achieved by different statistical techniques for robust analysis.

2 Data source and methodological framework

The study used number of promising socio-economic variables to determine the dynamic relationship between PPG factors and crime rate under the framework of an inverted U-shaped KC in a panel of 16 diversified countries, using system GMM estimator for the period of 1990–2014. The study used the following variables, i.e., crime rate (proxy by intentional homicides rate per 100,000 population), GINI index measures income inequality, poverty headcount ratio at $1.90 a day (2011 PPP) (% of total population), national estimates of unemployment in % of total labor force, education expenditures as % of GDP, per capita health expenditure in current US$, per capita income in constant 2005 US$, and trade openness as % of GDP. The samples of countries are presented in Table  7 in Appendix for ready reference. The data for the study are obtained from World Development Indicators published by World Bank ( 2015 ).

These countries are selected because of the devastating crime rate during the study time period. The recorded figures for Argentina crime rates about to 245% increase between the period of 1991 and 2007, while 2002 is considered the highest committed crime data recorded when the POV and INC_INEQ reached at their peak levels (Bouzat 2010 ). Brazil economy is working out for reduction of crime by focusing on three-point agenda, i.e., reduction in income disparity, to increase spending on education via an increase in enrollment of school dropout children, and to improve labor market conditionings. These three policies design to deter the crime rates in a given country (World Bank 2013 ). The robbery complaints largely increase since last two decades in Chile, which is being planned by controlling two action strategies, i.e., plan cuadrante and country security plan. Both the plan designed to restructured police force to reduce robbery and violence in a country (Vergara 2012 ). The rural China is suffered by high INC_INEQ that leads to higher crime rate (South China Monitoring Report 2015 ) while POV and INC_INEQ lead to crime and violent factor in Colombia (Gordon 2016 ). The socio-economic factors including low provision of education, health, high POV, and food challenges lead to increase crime in Indonesia (Pane 2017 ), while generating employment opportunities and increasing wage rate in Malaysia may be beneficial to reduce crime–POV nexus in a given country (Mulok et al. 2017 ). Mexican economy is suffered with high rate of homicides that negatively affect labor market outcomes, while country inhibits by increasing strict laws to diminish violence (Kato Vidal 2015 ). The safety situation in Morocco is cumbersome, as one of the country reports shows that an increased rate in crime is about to increase up to 23% in 2016 (OSAC 2017 ). The number of other factors remains visible in selected sample of panel of countries, including rural POV and social exclusion that is considered the main factor of socio-economic crisis in Poland (European Commission 2008 ); POV, unemployment, and INC_INEQ chiefly attributed to crime rate in South Africa (Bhorat et al. 2017 ); politics, democracy, and INC_INEQ arise conflicts in Thailand (Hewison 2014 ); corruption and high unemployment are the major conflicts in Tunisia (Saleh 2011 ); and Uruguay economy needs policy actions to reduce POV by investment in children education, modernizing rural sector, and balancing the gender gap (Thamma 2017 ). Thus, these facts about crime and POV in different countries put a focus to study crime–POV nexus under PPG framework in this study for robust evaluation. Figure  2 in Appendix shows the plots of the studied variables at level.

The study used the following non-linear equations to determine the dynamic relationship between PPG factors and crime rate in a panel of countries, i.e.,

where GDPPC indicates per capita GDP, GDPPC 2 indicates square of per capita GDP, GINI indicates Gini coefficient—income inequality, EDUEXP indicates education expenditures, HEXP indicates health expenditures, POVHCR indicates poverty headcount ratio, TOP indicates trade openness, UNEMP indicates unemployment, and CRIME indicates crime rate.

Equations ( 1 ) to ( 3 ) assessed the possible inverted U-shaped relationships between crime rate and PCI, between POVHCR and PCI, and between GINI and PCI, while Eq. ( 4 ) reviewed the PPG reforms across countries. Arellano and Bond ( 1991 ) developed the differenced GMM estimator, whom argued that the GMM estimator eliminates country effects and controls the possible endogeneity of explanatory variables using the appropriate instrumental list that evaluated by Sargan–Hansen test. The process further involves two-step GMM iterations with the time updated weights and adopted the weighting matrix by White period. The tests for autocorrelations by AR(1) and AR(2) and the Sargan test by Sargan–Hansen of over-identifying restrictions are presented for statistical reliability of the given models. The differenced GMM is superior to the 2SLS and system GMM, i.e., 2SLS regression estimator is used when the known endogeneity exists between the variables, which are handled by including the list of instrumental variables at their first lagged. Thus, the possible endogeneity problem is resolved accordingly. The system GMM further be used instead of 2SLS as if there are more than one endogenous issues exist in the model, which is unable to resolve through 2SLS estimator. Finally, the differenced GMM estimator is used as its estimated AR(1) and AR(2) bound values that would be helpful to encounter the issues of serial correlation and endogeneity problem accordingly.

Using the GMM estimator, the study verified different possibilities of KC, i.e., if the signs and magnitudes of \(\beta_{1} > 0\) and \(\beta_{2} < 0\) , than we may confirm the crime-induced KC, poverty-induced KC, and inequality-induced KC. The inverted U-shaped relationship between crime rate and PCI verified ‘crime-induced KC’, between POVHCR and PCI verified ‘POV-induced KC’, and inverted U-shaped relationship between GINI and PCI verified ‘inequality-induced KC’. On the other way around, if \(\beta_{1} < 0\) and \(\beta_{2} > 0\) , then we consider the U-shaped KC between crime rate and PCI, between POV and PCI, and between GINI and PCI, respectively. There are three other situations we may observe with the sign and magnitude of \(\beta_{1}\) and \(\beta_{2}\) , i.e., (i) \(\beta_{1} < 0\) and \(\beta_{2} = 0\) , (ii) \(\beta_{1} > 0\) and \(\beta_{2} = 0\) , and (iii) \(\beta_{1} = 0\) and \(\beta_{2} = 0\) , referred the monotonically decreasing function, monotonically increasing function, and flat/no relationship with the crime-PCI, poverty-PCI, and inequality-PCI in a panel of cross-sectional countries. The study further employed social accounting matrix by impulse response function (IRF) and variance decomposition analysis (VDA) in an inter-temporal relationship between the studied variables for a next 10-year period starting from 2015 to 2024. As it name implies, VDA explains the proportional variance in one variable caused by the proportional variance by the other variables in a vector autoregressive (VAR) system, while IRF traces the dynamic responses of a variable to innovations in other variables in the system. Both the techniques use the moving average representation of the original VAR system. Figure  1 shows the theoretical framework of the study to clearly outline the possible relationship between the stated variables.

Source: authors’ extraction

Research framework of the study.

Figure  1 shows the possible relationship between POV and crime rates in mediation of inequality, unemployment, and EG across countries. It is likelihood that POV increases inequality that leads to decrease in EG. The low-income growth further leads to increased unemployment, which causes high crime rates. This nexus is still rotated through crime rates that increase POV incidence across countries. The PPG process still works under the stated factors that need judicious income distribution to reduce crime rates.

The study further proceeds to evaluate the PPG reforms in a panel of selected countries. Kakwani and Pernia ( 2000 ) proposed an index of PPG called ‘PPG index’, which is evaluated by the growth elasticity and inequality elasticity with respect to POV. The same methodology is adopted in this study to assess the PPG and/or pro-rich growth reforms to assess the changes in the crime rate in a panel of countries. PPG defined as a state in which where the growth trickles down to the poor as compared to the non-poor. Poverty is largely affected by two main factors, i.e., higher growth rate may reduce the POV rates, while higher INC_INEQ reduces the impact of EG to reduce POV; therefore, the PPG index included the following mathematical illustrations, i.e.,

The study further assessed the pro-poorness of social expenditures and evaluates its impact to observe changes in IHR. The study shows the following mathematical illustrations that is extended from the scholarly work of Zaman and Khilji ( 2014 ); Kakwani and Pernia ( 2000 ) and Kakwani and Son ( 2004 ) i.e.,

where \(\alpha =\) 0, 1 and 2 indicate POVHCR, poverty gap and squared poverty gap, respectively, ‘P’ indicates FGT poverty measures, and ‘SOCIALEXP’ indicates social expenditures. Differentiating \(\eta_{\alpha }\) in Eq. ( 9 ) with respect to social expenditures gives more elaborated form of GEP, i.e.,

The elasticity of entire class of poverty measures \(P_{\alpha }\) with respect to Gini index is given by

which will be always positive only when \(S{\text{OCIALEXPE}} > z\) .Equations ( 10 ) and ( 11 ) are combined together to form TPE for all FGT poverty measures, i.e.,

or \(\delta_{\alpha } = \eta_{\alpha } + \xi_{\alpha }\) . Finally, pro-poorness of social expenditures estimated based on the following equation, i.e.,

Kakwani and Son ( 2004 ) presented the following bench mark applications to assess the pro-poor and/or anti-poor policies, i.e., the following value judgments regarding the PPG index ( \(\varphi\) ) are as follows, i.e.,

\(\varphi\)  < 0, growth is pro-rich or anti-poor,

0 <  \(\varphi\) \(\le\) 0.33, the process of PPG is considerable low,

0.33 <  \(\theta\) \(\le\) 0.66, the process of PPG is moderate,

0.66 <  \(\varphi\)  < 1.0, the process of EG considered as pro-poor, and

\(\varphi \ge\) 1.0, the process of EG is highly pro-poor.

The study utilized the PPG model for ready reference in this study.

This section presented the descriptive statistics in Table  1 , correlation matrix in Table  2 , dynamic system GMM estimates in Table  3 , IRF estimates in Table  4 , VDA estimates in Table  5 , while finally Table  6 shows the estimates for PPG in a panel of selected countries. Table  1 shows that GDPPC has a minimum value of US$ 199.350 and the maximum value of US$ 11257.600, with a mean and standard deviation (STD) value of US$ 4340.777 and US$ 2490.554, respectively. GINI has a minimum value of 25% and the maximum value of 64.790%, having an STD value of 8.580% with an average value of 45.095%. The minimum value of EDUEXP is about 0.998% of GDP and the maximum value of 7.657% of GDP, with an average value of 4.051% of GDP. The average value of HEXP per capita is about US$ 321.249 and a maximum value of US$ 1431.154, with an STD value of US$ 292.802. The maximum value of POVHCR is about 69% at US$1.90 a day with an average value of 12.394% at US$1.90 a day. The minimum value of trade is 13.753% of GDP and the maximum value of 220.407% of GDP, with an average value of 62.391% of GDP. The mean value for UNEMP is about 8.890% of total labor force with STD value of 6.010%. Finally, the minimum value of crime rate is about 0.439 per 100,000 inhabitants and the maximum value of 71.786 per 100,000 inhabitants, with an average value of 11.664 per 100,000 peoples. This exercise would be helpful to understand the basic descriptions of the studied variables in a panel of countries.

Figure  3 in Appendix shows the plots of the studied variables and found the stationary movement in the variables at their first difference. Table  2 presents the estimates of correlation matrix and found that GINI (i.e., r  = 0.264), EDUEXP ( r  = 0.243), HEXP ( r  = 0.730), TOP ( r  = 0.061), UNEMP (0.152) and CRIME ( r  = 0.031) have a positive correlation with the GDPPC, while POVHCR ( r  = − 0.599) significantly decreases GDPPC.

The results further reveal that GINI is affected by EDUEXP, HEXP, UNEMP and CRIME, while it considerably decreases by trade liberalization policies. EDUEXP, HEXP, PCI, TOP and UNEMP significantly decrease POVHCR, while crime rate has a positive correlation with the POVHCR. Finally, GINI have a greater magnitude, i.e., r  = 0.671, to influence CRIME, followed by UNEMP ( r  = 0.417), EDUEXP ( r  = 0.188), and POVHCR ( r  = 0.164) while trade liberalization policies support to decrease crime rates in a panel of countries. The study now proceeds to estimate the two-step system GMM for analyzing the functional relationship between socio-economic factors and crime rate. The results are presented in Table  3 .

The results of panel GMM show that GINI and UNEMP both have a significant and direct relationship with the CRIME, while TOP have an indirect relationship with CRIME in a panel of countries. The results imply that GINI and UNEMP are the main factors that increase CRIME, while trade liberalization policies have a supportive role to decrease crime rates across countries. Thorbecke and Charumilind ( 2002 ) evaluated the impact of income inequality on health, education, political conflict, and crime, and surveyed the different casual mechanism in between income inequality and its socio-economic impact across the globe. The policies have devised while reaching the conclusive relationships between them. Kennedy et al. ( 1998 ) concluded that social capital and income inequality are the powerful predictors of intentional homicides rate and violent crime in the US states. Altindag ( 2012 ) explored the long-run relationship between unemployment and crime rates in a country-specific panel dataset of Europe and found that unemployment significantly increases crime rates, while unemployment has a power predictor of exchange rate movements and industrial accident across the Europe. Menezes et al. ( 2013 ) confirmed the positive association between income inequality and criminality, as rational income distribution tends to decrease neighborhood homicides rate while it implies an increase in the intentional homicides rate in the surrounding neighborhoods.

In a second regression panel, the results confirmed the U-shaped relationship between POVHCR and GDPPC, as at initial level of EG, POV significantly declines, while at the later stages, this result is evaporated, as EG subsequently increases POVHCR that shows pro-rich federal policies across countries. The HEXP, however, significantly decreases POVHCR during the study time period. Dercon et al. ( 2012 ) investigated the relationship between chronic POV and rural EG in Ethiopia and argued that chronic POV is associated with the lack of education, physical assets and remoteness, while EG in terms of provide better roads and extension services may trickle down to the poor in a same way that the non-chronically poor benefited. Solinger and Hu ( 2012 ) examined the relationship between health, wealth and POV in urban China and found that wealthier cities prefer to allocate their considerable portion of savings for social assistance funds, while poorer places save the city money and work outside in a hope that the peoples would be better able to support themselves. Fosu ( 2015 ) examined the relationship between GIP triangle in sub-Saharan African countries and found that as a whole, South African countries lag behind the BICR (Brazil, India, China and Russia) group of countries; however, many of them in sub-Saharan African countries have outperformed India. The results further specified that PCI is the main predictor to reduce POV in sub-Saharan African countries; however, rational income distribution is a crucial challenge to reduce POV reduction through substantial growth reforms in a region. Kalichman et al. ( 2015 ) concluded that food poverty is associated with the multifaceted problems of health-related outcomes across the globe.

In a third regression panel, the results confirm an inverted U-shaped relationship between GDPPC and GINI that verified an inequality-induced KC in a panel of countries. The results imply that at initial level of economic development, GINI first increases and then decreases with the increased GDPPC across countries. CRIME, however, it is associated with the higher GINI during the studied time period. Kuznets ( 1955 ), Ahluwalia ( 1976 ), Deininger and Squire ( 1998 ), and others confirmed an inverted U-shaped relationship between INC_INEQ and PCI in different economic settings. Mo ( 2000 ) suggested different channelss to examine the possible impact of INC_INEQ on EG and found that ‘transfer channel’ exert the most important channel, while ‘human capital’ is the least important channel that negatively affects the rate of EG via INC_INEQ. Popa ( 2012 ) argued that health and education both are important predictors for EG, while POV and unemployment negatively correlated with the EG in Romania. Herzer and Vollmer ( 2012 ) confirmed the negative relationship between INC_INEQ and EG within the sample of developing countries, developed countries, democracies, non-democracies, and sample as a whole. In a similar line, Malinen ( 2012 ) confirmed the long-run equilibrium relationship between PCI and INC_INEQ and found that income inequality negatively affected the growth of developed countries.

The final regression shows that HEXP and TOP both significantly increase GDPPC, while POVHCR decreases the pace of EG, which merely be shown pro-rich federal policies in a panel of countries. Ranis et al. ( 2000 ) found that both the health and education expenditures lead to increased EG, while investment improves human development in a cross-country regression. Bloom et al. ( 2004 ) confirmed the positive connection between health and EG across the globe. Gyimah-Brempong and Wilson ( 2004 ) examined the possible effect of healthy human capital on PCI of sub-Saharan African and OECD countries and found the positive association between them in a panel of countries.

The statistical tests of the system GMM estimator confirmed the stability of the model by F-statistics, as empirically model is stable at 1% level of confidence interval. Sargan–Hansen test confirmed the instrumental validity at conventional levels for all cases estimated. Autocorrelations tests imply that except POVHCR model, the remaining three models including CRIME, GINI and GDPPC model confirmed the absence of first- and second-order serial correlation, and as a consequence, we verified our instruments are valid. As far as POVHCR model, we believed the results of Sargan–Hansen test of over identifying restrictions and AR(1) that is insignificant at 5% level, and confirmed the validity of instruments and absence of autocorrelation at first-order serial correlation. Table  4 shows the estimate of IRF for the next 10-year period starting from a year of 2015 to 2024.

The results show that the socio-economic factors have a mix result with the rate of crime, as POVHCR slightly increases with decreasing rate with the crime data, i.e., in the next coming years from 2016, 2018, 2019, and 2022, POVHCR exhibits a negative sign, while in the remaining years in between from 2015 to 2024, POVHCR increases crime rate. GINI will considerably increase crime rate from 2022 to 2024. UNEMP has a mixed result to either increase crime rate in one period while in the very next upcoming periods, it declines crime rate. Similar types of results been found with EDUEXP, HEXP and with the TOP; however, GDPPC will constantly increase the rate of crime in a panel of countries. In an inter-temporal relationship between POVHCR and other predictors, the results show that GDPPC would significantly decrease POVHCR for the next 10-year period; however, UNEMP, HEXP, and crime rate would considerably increase POVHCR. EDUEXP and TOP would support to reduce GINI for the next upcoming years, while remaining variables including crime rate, POV, UNEMP, HEXP, and GDPPC associated with an increased GINI across countries. The GDPPC will be influenced by crime rate, POVHCR, GINI, UNEMP, HEXP, and EDUEXP, while TOP would considerably to support GDPPC for the next 10-year time period. Figure  4 in Appendix shows the IRF estimates for the ready reference.

Table  5 shows the estimates of VDA and found that POVHCR will exert the largest share to influence crime rates, followed by GDPPC, TOP, HEXP, EDUEXP, GINI, and UNEMP. POVHCR would be affected by crime rate (i.e., 4.450%), UNEMP (1.751%), GDPPC (1.120%), GINI (1.043%), HEXP (0.639%), and EDUEXP (0.512%), and TOP (0.299%), respectively.

The results further reveal that GINI will affected by POVHCR, as it is explained by 7.680% variations to influence GINI for the next 10-year period. UNEMP, EDUEXP, and crime rate will subsequently influenced GDPPC about to 1.107%, 0.965%, and 0.312% respectively. The largest variance to explain UNEMP will be TOP, while the lowest variance to influence UNEMP will be GINI for the next 10-year period. Finally, GDPPC would largely influenced by HEXP, followed by UNEMP, CRIME, POVHCR, EDUEXP, TOP, and GINI for the period of 2015 to 2024. Figure  5 in Appendix shows the plots of the VDA for ready reference.

Finally, Table  6 presents the changes in crime rate by five different growth phases, i.e., phase 1: 1990–1994, phase 2: 1995–1999, phase 3: 2000–2004, phase 4: 2005–2009, and phase 5: 2010–2014. The results show that in the years 1990–1994, 1% increase in EG and INC_INEQ decrease POVHCR by − 0.023% and − 0.630%, which reduces TPE by − 0.629 percentage points. The PPG index surpassed the bench mark value of unity and confirmed the trickledown effect that facilitates the poor as compared to the non-poor. However, there is an overwhelming increase in the crime rate beside that the pro-poorness of EG, which indicate the need for substantial safety nets’ protection to the poor that escape out from this acute activities (Wang et al. 2017 ). In a second phase from 1995 to 1999, although EG decreases POVHCR by − 0.187; however, GINI has a greater share to increase POVHCR by 0.517% that ultimately increases TPE by 0.330%. This increase in the TPE turns to decrease PPG as 1.764, which shows anti-poor/pro-rich federal policies and low reforms for the poor that accompanied with the higher rates of crime in a panel of countries. The rest of the growth phases from 2000 to 2014 show anti-poor growth accompanied with the higher INC_INEQ and lower EG; however, crime rate decreases in the year 2000–2004 and 2010–2014 besides that the growth process is anti-poor across countries. The policies should be formulated in a way to aligned crime rate with the PPG reforms across countries (Vellala et al. 2018 ).

The results of PPE index confirmed an anti-poor growth from 1990 to 2004, while at the subsequent years from 2005 to 2014, education growth rate subsequently benefited the poor as compared to the non-poor, i.e., PPE index exceeds the bench mark value of unity. Crime rate is increasing from 1990 to 1999, and from 2005 to 2009, while it decreases the crime rate for the years 2000–2004 and 2010–2014. The good sign of recovery has been visible for the years 2010–2014 where the PPE growth supports to decrease crime rate in a panel of selected countries. Finally, the PPH index confirmed two PPG phases, i.e., from 1990 to 1994, and 2010 to 2014 in which crime rate increases for the former years and decreases in the later years. The remaining health phases from 1995 to 2009 show anti-poor health index, while crime rate is still increasing during the years from 1995 to 1999 and 2005 to 2009, and decreasing for the period 2000–2004. The results emphasized the need to integrate PPG index with the crime rate, as PPG reforms are helpful to reduce humans’ costs by increasing EG and social expenditures, and providing judicious income distribution to escape out from POV and vulnerability across the globe (Musavengane et al. 2019 ).

From the overall results, we come to the conclusion that social spending on education and health is imperative to reduce crime incidence, while it further translated a positive impact on POV and inequality reduction across countries (Hinton 2016 ). EG is a vital factor to reduce POV; however, it is not a sufficient condition under higher INC_INEQ (Dudzevičiūtė and Prakapienė 2018 ). INC_INEQ and unemployment rate both are negatively correlated with crime rates; however, it may be reduced by judicious income distribution and increases social spending across countries (Costantini et al. 2018 ). Trade liberalization policies reduce incidence of crime rates and improve country’s PCI, which enforce the need to capitalize domestic exports by expanding local industries. Thus, the United Nations SDGs would be achieved by its implication in the countries perspectives (Dix-Carneiro et al. 2018 ). The study achieved the research objectives by its theoretical and empirical contribution, which seems challenge for the developmental experts to devise policies toward more pro-growth and PPG.

4 Conclusions and policy recommendations

This study investigated the dynamic relationship between socio-economic factors and crime rate to assess PPG reforms for reducing crime rate in a panel of 16 diversified countries, using a time series data from 1990–2014. The study used PCI and square PCI in relation with crime rate, POVHCR, and GINI to evaluate crime-induced KC, poverty-induced KC and inequality-induced KC, while PPG index assesses the federal growth reforms regarding healthcare provision, education and wealth to escape out from POV and violence. The results show that GINI and UNEMP are the main predictors that have a devastating impact to increase crime rate. Trade liberalization policies are helpful to reduce crime rate and increase PCI. Healthcare expenditures decrease POVHCR and amplify EG. The EG is affected by POVHCR, which requires strong policy framework to devise PPG approach in a panel of selected countries. The study failed to establish crime-induced KC and poverty-induced KC, while the study confirmed an inequality-induced KC. The results of IRF reveal that PCI would considerably increase crime rate, while crime rate influenced GINI and PCI for the next 10-year period. The estimates of VDA show that POVHCR explained the greater share to influence crime rates, while reverse is true in case of POVHCR. The study divided the studied time period into five growth phases 1990–1994, 1995–1999, 2000–2004, 2005–2009, and 2010–2014 to assess PPG, PPH, and PPE reforms and observe the changes in crime rates. The results show that there is an only period from 1990 to 1994 that shows PPG, while crime rate is still increasing in that period; however, in the years 2000–2004, and 2010–2014, crime rate decreases without favoring the growth to the poor. PPE and PPH assessment confirmed the reduction in the crime rates for the years 2010–2014. The overall results confirmed the strong correlation between socio-economic factors and crime rates to purse the pro-poorness of government policies across countries. The overall results emphasized the need of strong policy framework to aligned PPG policies with the reduction in crime rate across the globe. The study proposed the following policy recommendations, i.e.,

Education, health and wealth are the strong predictors of reducing crime rates and achieving PPG, thus it should be aligned with inclusive trade policies to reduce human cost in terms of decreasing chronic poverty and violence/crime.

The policies should be formulated to strengthen the pro-poorness of social expenditures that would be helpful to reduce an overwhelming impact of crime rate in a panel of countries.

GIP triangle is mostly viewed as a pro-poor package to reduce the vicious cycle of poverty; however, there is a strong need to include some other social factors including unemployment, violence, crime, etc., which is mostly charged due to increase in poverty and unequal distribution of income across the globe. The policies should devise to observe the positive change in lessen the crime rate by PPG reforms in a panel of selected countries.

The significant implication of the Kuznets’ work should be extended to the some other unexplored factors especially for crime rate that would be traced out by the pro-poor agenda and pro-growth reforms.

There is a need to align the positivity of judicious income distribution with the broad-based economic growth that would be helpful to reduce poverty and crime rate across countries.

The result although not supported the ‘parabola’ relationship between income and crime rates; however, it confirmed the U-shaped relationship between income and poverty. The economic implication is that income is not the sole contributor to increase crime rates while poverty exacerbates violent crimes across countries. There is a high need to develop a mechanism through which poverty incidence can be reduced, which would ultimately lead to decreased crime rates. The improvement in the labor market structure, judicious income distribution, and providing social safety nets are the desirable strategies to reduce crime rates and poverty incidence across countries, and

The results supported parabola relationship between economic growth and inequality, which gives a clear indication to improve income distribution channel for reducing poverty and crime rates at global scale.

These seven policies would give strong alignment to improve social infrastructure for managing crime through equitable justice and PPG process.

Availability of data and materials

The data are freely available on World Development Indicator, published by World Bank on given URL ID: https://datacatalog.worldbank.org/dataset/world-development-indicators .

Adeleye NB (2014). The determinants of income inequality and the relationship to crime. Unpublished dissertation, University of Sussex, UK. https://www.researchgate.net/profile/Ngozi_Adeleye2/publication/276410308_The_Determinants_of_Income_Inequality_and_the_Relationship_to_Crime/links/5558c9f808aeaaff3bf98a45.pdf . Accessed 6 Jan 2016

Ahluwalia MS (1976) Income distribution and development: some stylized facts. Am Econ Rev 66(2):128–135

Google Scholar  

Ahmed K, Mahalik MK, Shahbaz M (2016) Dynamics between economic growth, labor, capital and natural resource abundance in Iran: an application of the combined cointegration approach. Res Policy 49:213–221

Article   Google Scholar  

Altindag DT (2012) Crime and unemployment: evidence from Europe. Int Rev Law Econ 32(1):145–157

Arellano M, Bond SR (1991) Some tests of specification of panel data: monte Carlo evidence and an application to employment equations. Rev Econ Stud 58:277–297

Berens S, Gelepithis M (2018) Welfare state structure, inequality, and public attitudes towards progressive taxation. Socio Econ Rev. https://doi.org/10.1093/ser/mwx063

Bhorat H, Thornton A, Van der Zee K (2017). Socio-economic determinants of crime in South Africa: an empirical assessment. DPRU Working Paper 201704. DPRU, University of Cape Town, Rondebosch

Bloom DE, Canning D, Sevilla J (2004) The effect of health on economic growth: a production function approach. World Dev 32(1):1–13

Bourguignon F (2000). Crime, violence and inequitable development. Annual World Bank Conference on development economics 1999, pp. 199–220

Bouzat G (2010) Inequality, crime, and security in Argentina. SELA (Seminario en Latinoamérica de Teoría Constitucional y Política) Papers. Paper 91. http://digitalcommons.law.yale.edu/yls_sela/91 . Accessed 6 October 2018

Chu DC, Tusalem RF (2013) The role of the state on cross-national homicide rates. Int Crim Justice Rev 23(3):252–279

Costantini M, Meco I, Paradiso A (2018) Do inequality, unemployment and deterrence affect crime over the long run? Reg Stud 52(4):558–571

Dalberis R (2015). Extreme levels of poverty and inequality may lead to equally high levels of social conflict and crime. Unpublished dissertation, CUNY Academic Works, New York. http://academicworks.cuny.edu/cc_etds_theses/346/ . Accessed 6 Jan 2016

Deininger K, Squire L (1998) New ways of looking at old issues: inequality and growth. J Dev Econ 57(2):259–287

Dercon S, Hoddinott J, Woldehanna T (2012) Growth and chronic poverty: evidence from rural communities in Ethiopia. J Dev Stud 48(2):238–253

Dix-Carneiro R, Soares RR, Ulyssea G (2018) Economic shocks and crime: evidence from the Brazilian trade liberalization. Am Econ J Appl Econ 10(4):158–195

Dreze J, Khera R (2000) Crime, gender, and society in India: insights from homicide data. Popul Dev Rev 26(2):335–352

Dudzevičiūtė G, Prakapienė D (2018) Investigation of the economic growth, poverty and inequality inter-linkages in the European Union countries. J Secur Sust Issues 7:839–854

Duque M, McKnight A (2019) Understanding the relationship between inequalities and poverty: mechanisms associated with crime, the legal system and punitive sanctions. LIP Paper, 6, CASE/215, Centre for Analysis of Social Exclusion, London School of Economics, London, UK. http://sticerd.lse.ac.uk/dps/case/cp/casepaper215.pdf . Accessed 8 March 2020

Enamorado T, López-Calva LF, Rodríguez-Castelán C, Winkler H (2016) Income inequality and violent crime: evidence from Mexico’s drug war. J Dev Econ 120:128–143

European Commission (2008 ). Poverty and social exclusion in rural areas—final report Annex I—Country Studies-Poland. European commission, Brussels

Fosu AK (2015) Growth, inequality and poverty in Sub-Saharan Africa: recent progress in a global context. Oxford Dev Stud 43(1):44–59

Gordon E (2016) Poverty, crime, and conflicts: socio-economic inequalities and prospects for peace in Colombia. Centre for security governance. http://secgovcentre.org/2016/10/poverty-crime-and-conflict-socio-economic-inequalities-and-the-prospects-for-peace-in-colombia/ . Accessed 6 October 2018

Gyimah-Brempong K, Wilson M (2004) Health human capital and economic growth in Sub-Saharan African and OECD countries. Q Rev Econ Finance 44(2):296–320

Harris G, Vermaak C (2015) Economic inequality as a source of interpersonal violence: evidence from Sub-Saharan Africa and South Africa. South Afr J Econ Manag Sci 18(1):45–57

Heinemann, Verner D (2006) Crime and violence in development: a literature review of Latin America and the Caribbean. The World Bank, Policy Research Working Papers, Washington D.C

Herzer D, Vollmer S (2012) Inequality and growth: evidence from panel cointegration. J Econ Inequality 10(4):489–503

Hewison K (2014) Considerations on inequality and politics in Thailand. Democratization 21(5):846–866

Hinton E (2016) From the war on poverty to the war on crime: the making of mass incarceration in America. Harvard University Press, Cambridge

Book   Google Scholar  

Imran M, Hosen M, Chowdhury MAF (2018) Does poverty lead to crime? Evidence from the United States of America. Int J Soc Econ 45(10):1424–1438

Jacobs D, Richardson AM (2008) Economic inequality and homicide in the developed nations from 1975 to 1995. Homicide Stud 12(1):28–45

Kakwani N, Pernia EM (2000) What is pro-poor growth? Asian Dev Rev 18(1):1–16

Kakwani N, Son HH (2004) Pro-poor growth: concepts and measurement with country case studies. Pakistan Dev Rev 42(4 Part I):417–444

Kalichman SC, Hernandez D, Kegler C, Cherry C, Kalichman MO, Grebler T (2015) Dimensions of poverty and health outcomes among people living with HIV infection: limited resources and competing needs. J Community Health 40(4):702–708

Kato Vidal EL (2015) Violence in Mexico: an economic rationale of crime and its impacts. EconoQuantum 12(2):93–108

Kelly M (2000) Inequality and crime. Rev Econ Stat 82(4):530–539

Kennedy BP, Kawachi I, Prothrow-Stith D, Lochner K, Gupta V (1998) Social capital, income inequality, and firearm violent crime. Soc Sci Med 47(1):7–17

Khan N, Ahmed J, Nawaz M, Zaman K (2015) The socio-economic determinants of crime in Pakistan: new evidence on an old debate. Arab Econ Business J 10(2):73–81

Khan HUR, Khan A, Zaman K, Nabi AA, Hishan SS, Islam T (2017) Gender discrimination in education, health, and labour market: a voice for equality. Qual Quant 51(5):2245–2266

Khan HUR, Zaman K, Yousaf SU, Shoukry AM, Gani S, Sharkawy MA (2019a) Socio-economic and environmental factors influenced pro-poor growth process: new development triangle. Environ Sci Pollut Res 26(28):29157–29172

Khan HUR, Nassani AA, Aldakhil AM, Abro MMQ, Islam T, Zaman K (2019b) Pro-poor growth and sustainable development framework: evidence from two step GMM estimator. J Cleaner Prod 206:767–784

Kousar R, Rais SI, Mansoor A, Zaman K, Shah STH, Ejaz S (2019) The impact of foreign remittances and financial development on poverty and income inequality in Pakistan: evidence from ARDL-bounds testing approach. J Asian Finance Econ Business 6(1):71–81

Kuznets S (1955) Economic growth and income inequality. Am Econ Rev 45(1):1–28

Ling CH, Ahmed K, Muhamad R, Shahbaz M, Loganathan N (2017) Testing the social cost of rapid economic development in Malaysia: the effect of trade on life expectancy. Soc Indic Res 130(3):1005–1023

Liu Y, Fullerton TM, Ashby NJ (2013) Assessing the impacts of labor market and deterrence variables on crime rates in Mexico. Contemp Econ Policy 31(4):669–690

Malinen T (2012) Estimating the long-run relationship between income inequality and economic development. Empirical Econ 42(1):209–233

Menezes T, Silveira-Neto R, Monteiro C, Ratton JL (2013) Spatial correlation between homicide rates and inequality: evidence from urban neighborhoods. Econ Lett 120(1):97–99

Mo PH (2000) Income inequality and economic growth. Kyklos 53(3):293–315

Mukherjee S (2019) Crime and Social Deprivation across States in India–insights from a Panel Data Discourse on Social Sustainability’. The Impact of Global Terrorism on Economic and Political Development. Emerald Publishing Limited, pp. 249–265

Mulok D, Kogid M, Lily J, Asid R (2017) The relationship between crime and economic growth in Malaysia: re-examine using bound test approach. Malays J Business Econ (MJBE) 3(1):15–26

Musavengane R, Siakwah P, Leonard L (2019) “Does the poor matter” in pro-poor driven sub-Saharan African cities? towards progressive and inclusive pro-poor tourism. Int J Tourism Cities 5(3):392–411

Neumayer E (2003) Good policy can lower violent crime: evidence from a cross-national panel of homicide rates, 1980–97. J Peace Res 40(6):619–640

OSAC (2017) Morocco 2017 crime & safety report: rabat. Overseas Security Advisory Council, Washington DC

Ouimet M (2012) A world of homicides the effect of economic development, income inequality, and excess infant mortality on the homicide rate for 165 countries in 2010. Homicide Stud 16(3):238–258

Pane H (2017) The Social Problems of National Poverty and Criminality in Indonesia. Int J Soc Sci Human Invention 4(8):3834–3836

Piatkowska SJ (2020) Poverty, inequality, and suicide rates: a Cross-National Assessment of the Durkheim Theory and the Stream Analogy of Lethal Violence. Soc Q. 19:1–26

Popa AM (2012) The impact of social factors on economic growth: empirical evidence for Romania and European Union countries. Rom J Fiscal Policy (RJFP) 3(2):1–16

Pridemore WA (2011) Poverty matters: a reassessment of the inequality–homicide relationship in cross-national studies. Br J Criminol 51(5):739–772

Ranis G, Stewart F, Ramirez A (2000) Economic growth and human development. World Dev 28(2):197–219

Saasa OS (2018) Poverty profile in sub-Saharan Africa: the challenge of addressing an elusive problem. Contested terrains and constructed categories. Routledge, Abingdon, pp 105–116

Sachsida A, de Mendonça MJC, Loureiro PR, Gutierrez MBS (2010) Inequality and criminality revisited: further evidence from Brazil. Empirical Econ 39(1):93–109

Saleh B (2011) Tunisia’s economic medicine, poverty and unemployment. https://www.pambazuka.org/governance/tunisias-economic-medicine-poverty-and-unemployment . Accessed 6 October 2018

Solinger DJ, Hu Y (2012) Welfare, wealth and poverty in urban China: the Dibao and its differential disbursement. China Q 211:741–764

South China Monitoring Report (2015). Big trouble in rural China: data reveals greater the wealth gap the higher the crime rate, and Hong Kong is feeling the effects. Law and crime section. https://www.scmp.com/news/hong-kong/law-crime/article/1876931/big-trouble-rural-china-data-reveals-greater-wealth-gap . Accessed 6 October 2018

Stamatel JP (2016) Democratic cultural values as predictors of cross-national homicide variation in Europe. Homicide Stud 20(3):239–256

Thamman B (2017) Causes of poverty in Uruguay. The Borgen project. https://borgenproject.org/causes-of-poverty-in-uruguay/ . Accessed 6 October 2018

Thorbecke E, Charumilind C (2002) Economic inequality and its socioeconomic impact. World Dev 30(9):1477–1495

Ulriksen MS (2012) Questioning the pro-poor agenda: examining the links between social protection and poverty. Dev Policy Rev 30(3):261–281

Vellala PS, Madala MK, Chattopadhyay U (2018) Econometric analysis of growth inclusiveness in India: evidence from cross-sectional data. Advances in finance & applied economics. Springer, Singapore, pp 19–38

Vergara R (2012) Crime prevention programs: evidence from CHILE. Dev Econ 50(1):1–24

Wang C (2017) Which dimension of income distribution drives crime? Evidence from the People’s Republic of China. Asian Development Bank, Working paper no: 704, Tokyo, Japan. https://www.adb.org/publications/which-dimension-income-distribution-drives-crime-prc . Accessed 6 October 2018

Wang H, Yao H, Kifer D, Graif C, Li Z (2017) Non-stationary model for crime rate inference using modern urban data. IEEE Trans Big Data 5(2):180–194

World Bank (2013). Brazil Fights Crime while Bringing Development to the Favelas. http://www.worldbank.org/en/news/feature/2013/03/21/brazil-crime-violence-favela . Accessed 6 October 2018

World Bank (2015) World Development Indicators. World Bank, Washington DC

Zaman K (2018) Crime-poverty nexus: an intellectual survey. Forensic Res Criminol Int J 6(5):327–329

Zaman K, Khilji BA (2014) A note on pro-poor social expenditures. Qual Quant 48(4):2121–2154

Zaman K, Usman B, Sheikh SM, Khan A, Kosnin ABM, Rosman ASB, Hishan SS (2019) Managing crime through quality education: a model of justice. Sci Justice 59(6):597–605

Zaman K, Al-Ghazali BM, Khan A, Rosman ASB, Sriyanto S, Hishan SS, Bakar ZA (2020) Pooled mean group estimation for growth, inequality, and poverty triangle: evidence from 124 Countries. J Poverty 24(3):222–240

Download references

Acknowledgements

The authors are thankful for King Saud university research project number (RSP-2019/87) for funding the study. The authors are indebted to the editor and reviewers for constructive comments that have helped to improve the quality of the manuscript.

Researchers Supporting Project number (RSP-2019/87), King Saud University, Riyadh, Saudi Arabia.

Author information

Authors and affiliations.

Department of Public Administration, Xi’an University of Architecture and Technology, Xi’an, China

Muhammad Khalid Anser

Higher Education Department Khyber Pakhtunkhwa, Government College of Management Sciences, Abbottabad, Pakistan

Zahid Yousaf

Department of Management, College of Business Administration, King Saud University, P.O. Box 71115, Riyadh, 11587, Saudi Arabia

Abdelmohsen A. Nassani & Saad M. Alotaibi

Department of Management, Aleppo University, Aleppo, Syria

Ahmad Kabbani

Department of Economics, University of Wah, Quaid Avenue, Wah Cantt, Pakistan

Khalid Zaman

You can also search for this author in PubMed   Google Scholar

Contributions

The collection, processing and analysis of the data were carried out by the authors. All authors read and approved the final manuscript

Corresponding author

Correspondence to Khalid Zaman .

Ethics declarations

Competing interests.

The authors declare that they have no conflict of interest.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

See Table  7 , Figs.  2 , 3 , 4 and 5 .

Source: World Bank ( 2015 )

Data trend at level.

Source: World Bank ( 2015 ). ‘D’ indicates first difference

Data trend at first differenced

Source: authors’ estimation. Note: ‘D’ shows first difference, while ‘LOG’ represents natural logarithm

Plots of IRF.

VDA Estimates.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Anser, M.K., Yousaf, Z., Nassani, A.A. et al. Dynamic linkages between poverty, inequality, crime, and social expenditures in a panel of 16 countries: two-step GMM estimates. Economic Structures 9 , 43 (2020). https://doi.org/10.1186/s40008-020-00220-6

Download citation

Received : 03 February 2020

Revised : 09 March 2020

Accepted : 29 May 2020

Published : 05 June 2020

DOI : https://doi.org/10.1186/s40008-020-00220-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Pro-poor growth
• Income inequality
• Quality education
• Healthcare expenditures
• System panel GMM

Jel classification

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 21 January 2021

Why do inequality and deprivation produce high crime and low trust?

• Benoît De Courson 1 , 2 &
• Daniel Nettle 2  

Scientific Reports volume  11 , Article number:  1937 ( 2021 ) Cite this article

60k Accesses

22 Citations

355 Altmetric

Metrics details

• Human behaviour

Humans sometimes cooperate to mutual advantage, and sometimes exploit one another. In industrialised societies, the prevalence of exploitation, in the form of crime, is related to the distribution of economic resources: more unequal societies tend to have higher crime, as well as lower social trust. We created a model of cooperation and exploitation to explore why this should be. Distinctively, our model features a desperation threshold, a level of resources below which it is extremely damaging to fall. Agents do not belong to fixed types, but condition their behaviour on their current resource level and the behaviour in the population around them. We show that the optimal action for individuals who are close to the desperation threshold is to exploit others. This remains true even in the presence of severe and probable punishment for exploitation, since successful exploitation is the quickest route out of desperation, whereas being punished does not make already desperate states much worse. Simulated populations with a sufficiently unequal distribution of resources rapidly evolve an equilibrium of low trust and zero cooperation: desperate individuals try to exploit, and non-desperate individuals avoid interaction altogether. Making the distribution of resources more equal or increasing social mobility is generally effective in producing a high cooperation, high trust equilibrium; increasing punishment severity is not.

Similar content being viewed by others

Worldwide divergence of values

Network of large pedigrees reveals social practices of Avar communities

Strategy evolution on higher-order networks

Introduction.

Humans are often described as an unusually cooperative or ‘ultrasocial’ species 1 . The truth is more complex: humans from the same society can cooperate for mutual benefit; or they can simply co-exist; or they can actively exploit one another, as in, for example, crime. A theory of human sociality should ideally predict what mix of these alternatives will emerge under which circumstances. Comparing across industrialised societies, higher inequality—greater dispersion in the distribution of economic resources across individuals—is associated with higher crime and lower social trust 2 , 3 , 4 , 5 , 6 , 7 . These associations appear empirically robust, and meet epidemiological criteria for being considered causal 8 . However, the nature of the causal mechanisms is still debated. The effects of inequality are macroscopic phenomena, seen most clearly by comparing aggregates such as countries or states. It is their micro-foundations in individual psychology and behaviour that still require clarification.

There are, broadly, two classes of explanation for how inequality, a population-level phenomenon, could influence individual-level outcomes like crime or trust. The first class of explanation is compositional: in more unequal societies, the least fortunate individuals are absolutely worse off than in more equal societies of the same average wealth, exactly because the dispersion either side of the average is greater. Some individuals are also better off too, at the other end of the distribution, but if there is any non-linearity in the function relating individual resources to outcomes—if for example the poor becoming absolutely poorer has a larger effect on their propensity to offend than the rich becoming absolutely richer has on theirs—this can still change outcome prevalence in the population 9 , 10 , 11 . In line with compositional explanations, across US counties, the association between inequality and rate of property crime is fully mediated by the prevalence of poverty, which is higher in more unequal counties 2 . Moreover, changes in rates over time track changes in the economic prospects of people at the bottom end of the socioeconomic distribution 12 , 13 . The second class of explanation is psychosocial: individuals perceive the magnitude of social differentials in the society around them, and this affects their state of mind, increasing competitiveness, anxiety and self-serving individualism 8 , 14 . In this paper, we develop an explanatory model for why greater inequality should produce higher crime and lower social trust. Our model provides a bridge between compositional and psychosocial explanations. Its explanation for the inequality-crime association is compositional: individuals offend when their own absolute level of resources is desperately low, and the effect of increasing inequality is to make such desperation more prevalent. On the other hand, the model’s explanation for the inequality-trust association is more psychosocial: all individuals in high-inequality populations end up trusting less, regardless of their personal resource levels.

To provide a micro-foundation in individual behaviour for the macro-level effects of inequality on crime, we must start from explanations for why individuals commit crimes. Economic 15 , 16 and behavioural-ecological 17 approaches see offending as a strategic response to specific patterns of incentive. Economic models predict that offending should be more attractive when the payoffs from legitimate activity are low. This principle successfully explains variation in offending behaviour both within and between societies 12 , 16 . It can also explain the relationship between crime levels and inequality, in compositional manner, because unequal societies produce poorer legitimate opportunities for people at the lower end of the socioeconomic spectrum 2 . However, these models are generally taken to predict that making punishments for crime more severe should reduce the prevalence of offending, because harsher punishment should reduce the expected utility associated with the criminal option. Empirical evidence, though, does not clearly support the hypothesis that increasing punishment severity reduces offending 18 , 19 . There is more evidence for a deterrent effect of increased probability of punishment, though even this effect may be modest 18 , 19 .

Becker 15 pointed out that the puzzle of the weak deterrent effect of punishment severity would be solved if offenders were risk-preferring. The decision to offend is risky in that it has either a large positive payoff (if not caught) or a large negative one (if caught and punished). An individual who prefers risk might thus choose to offend even if the expected utility of offending is negative due to a possible severe punishment. Thus, the question becomes: why would some people—those who commit crime—prefer risk, when people are usually averse to it? To address this question, our model incorporates features of classic risk-sensitive foraging theory from behavioural ecology 20 (for a review in the context of human behaviour, see Ref. 21 ). Risk-sensitive foraging models incorporate a desperation threshold: a level of resources below which it is disastrous to fall, in the foraging case because of starvation. The models show that individuals in sufficient imminent danger of falling below this threshold ought to become risk-preferring. If a risky option is successful, it will allow them to leap back over the threshold; and if not, their prospects will be no more dire than they were anyway. Our model is novel in explicitly incorporating a desperation threshold into decisions about whether to cooperate (analogous in our model to participating in legitimate economic activity) or exploit others (analogous to committing an acquisitive crime).

The desperation threshold is the major theoretical innovation of our model. We justify its inclusion on multiple grounds. First, the ultimate currency in our model is fitness, a quantity with a natural biological interpretation that must necessarily be zero if the individual lacks the minimal resources to subsist and function socially. Thus, it is reasonable that expected fitness should be related to resource levels, but not linearly: there should be a point where, as resources deplete, expected fitness rapidly declines to zero. Our threshold assumption produces exactly this type of function (see Supplementary Sect.  2.1 , Supplementary Fig. S1 ). Second, in experimental games where gaining a payoff is subject to a threshold, people do switch to risk-proneness when in danger of falling below the threshold, as risk-sensitive foraging theory predicts 22 . Although this does not show that such thresholds are widespread or important in real life, it does show that people intuitively understand their implications when they are faced with them, and respond accordingly. Third, there are ethnographic descriptions of ‘disaster levels’, ‘crisis levels’, or ‘edges’ that affect the risk attitudes of people facing poverty 23 , 24 . For example, writing on Southeast Asia, Scott 23 describes the spectre of a “subsistence crisis level—perhaps a ‘danger zone’ rather than a ‘level’ would be more accurate…a threshold below which the qualitative deterioration in subsistence, security and status is massive and painful” (p. 17), as an ever-present factor in people’s decisions. Thus, including a desperation threshold is a simple but potentially powerful innovation into models of cooperation and exploitation, with potential to generate new insights.

In our model, agents repeatedly decide between three actions: foraging alone, foraging cooperatively, or exploiting a cooperative group. Foraging cooperatively is analogous to legitimate economic activity, and exploitation is analgous to acquisitive crime. Agents have variable levels of resources, and their behaviour is state-dependent. That is, rather than having a fixed strategy of always cooperating or always exploiting, each agent, at each interaction, selects a behaviour based on their current level of resources, the behaviour of others in the surrounding population, and background parameters such as the probability and severity of punishment, and the likelihood of resources improving through other means. Agents seek to maximize fitness. We assume that fitness is positively related to resource levels, but that there is a threshold, a critically low level of resources below which there is an immediate fitness penalty for falling. Our investigation of the model has two stages. We first compute the optimal action policy an individual should follow; that is, the optimal action to select for every possible combination of the situational variables. Second, we simulate populations of individuals all following the optimal action policies, to predict population-level outcomes for different initial resource distributions.

To explain the model in more detail, at each time point t in an indefinitely long sequence of time steps (where one time step is one economic interaction), agents have a current level of resources s. They can take one of three actions. Foraging alone costs x units of resources and is also guaranteed to return x. Thus, foraging alone is sufficient to maintain the agent but creates no increase in resources. It is also safe from exploitation, as we conceptualise it as involving minimal interaction with others. Alternatively, agents can team up with n-1 others to cooperate . As long as no other group member exploits, cooperation is mutually beneficial, costing x units but producing a payoff of \(\alpha x \left( {\alpha > 1} \right)\) to each group member. Finally, agents can exploit : join a cooperating group and try to selfishly divert the resources produced therein. If this exploitation is successful, they obtain a large reward β, but if they fail, they receive a punishment π. The probability of being punished is γ. The punishment is not administered by peers: we assume that there is a central punitive institution in place, and both the size and probability of punishment are exogenous. In our default case, the expected payoff for exploitation is zero (i.e. \(\left( {1 - \gamma } \right)\beta = \gamma \pi\) ), making exploitation no better than foraging alone on average, and worse than cooperating. However, the reward for a successful exploitation, β, is the largest payoff available to the agent in any single time step.

At every time step, each agent’s resource level is updated according to the outcomes of their action. In addition, resource levels change by a disturbance term controlled by a parameter r , such that the mean and variance of population resources are unchanged, but the temporal autocorrelation of agents’ resource levels is only 1 −  r . If r is high, individuals whose current resources are low can expect they will be higher in the future and vice versa, because of regression to the mean. If \(r = 0\) , resources will never change other than by the agent’s actions. We consider r a measure of social mobility due to causes other than choice of actions.

In the first stage, we use stochastic dynamic programming 25 , 26 to compute the optimal action policy. Fitness is a positive linear function of expected resource level s in the future. However, in computing the fitness payoffs of each action, we also penalize, by a fixed amount, any action that leaves the agent below a desperation threshold in the next time step (arbitrarily, we set this threshold at s  = 0). The optimal action policy identifies which one of the three actions is favoured for every possible combination of the factors that impinge on the agent. These include both their own current resource state s , and features of their social world, such as the severity of punishment π, the probability of punishment γ, and the level of social mobility r. A critical variable that enters into the computation of the optimal action is the probability that any cooperating group in the population will contain someone who exploits. We denote this probability p . We can think of 1 −  p as an index of the trustworthiness of the surrounding population. Computing the optimal policy effectively allows us to ask: under what circumstances should an individual forage alone, cooperate, or exploit?

In the second stage, we simulate populations of agents all following the optimal policies computed in the first stage. We can vary the starting distributions of resources (their mean and dispersion), as well as other parameters such as social mobility and the probability and severity of punishment. During the simulation stage, each agent forms an estimate of 1 −  p , the trustworthiness of others, through observing the behaviour of a randomly-selected subset of other individuals. We refer to these estimates as the agents’ social trust, since social trust is defined as the generalized expectation that others will behave well 27 . Social trust updates at the end of each time step. Agents’ social trust values are unbiased estimates of the current trustworthiness of the surrounding population, but they are not precise, because they are based on only a finite sample of other population members. The simulation stage, allows us to ask: what are the predicted temporal dynamics of behaviour, and of social trust, in populations with different starting distributions of resources, different levels of social mobility, and different punishments for exploitation?

Each of the three actions is optimal in a different region of the space formed by current resources s and the trustworthiness of others 1 −  p (Fig.  1 a). Below a critical value of s , agents should always exploit, regardless of trustworthiness . In the default case, this critical value is in the vicinity of the desperation threshold, though it can be lower or higher depending on the value of other parameters. With our default values, exploitation will not, on average, make the agent’s resource state any better in subsequent time steps. However, there is a large advantage to getting above the threshold in the next time step, and there is a region of the resource continuum where exploitation is the only action that can achieve this in one go (intuitively, it is the quickest way to ‘get one’s head above water’). Where s is above the critical value, cooperation is optimal as long as the trustworthiness of the surrounding population is sufficiently high. However, if trustworthiness is too low, the likelihood of getting exploited makes cooperation worse than foraging alone. The shape of the frontier between cooperation and foraging alone is complex when resources are close to the desperation threshold. This is because cooperation and foraging alone also differ in riskiness; foraging alone is risk-free, but cooperation carries a risk of being exploited that depends on trustworthiness. Just above the exploitation zone, there is a small region where cooperation is favoured even at low trustworthiness, since one successful cooperation would be enough to hurdle back over the threshold, but foraging alone would not. Just above this is a zone where foraging alone is favoured even at high trustworthiness; here the agent will be above the threshold in the next time period unless they are a victim of exploitation, which makes them averse to taking the risk of cooperating.

Optimal actions as a function of the individual’s current resources s and the trustworthiness of the surrounding population, 1 −  p . ( A ) All parameters at their default values. This includes: α = 1.2, r = 0.1, π = 10, and γ = 1/3 (see Table 1 for a full list). ( B ) Effect of altering the efficiency of cooperation α to be either lower (1.05) or higher (1.30) than ( A ). Other parameter values are as for ( A ). ( C ) Effects of varying social mobility, to be either high (r = 0.8), or complete (r = 1.0; i.e. resource levels in this time period have no continuity at all into the next). Other parameter values are as for ( A ). ( D ) Effect of increasing the severity of punishment for exploiters to π = 15 and π = 20. Other parameter values are as for ( A ). ( E ). Effects of altering the probability of punishment for exploiters to γ = 2/3 and γ = 9/10. Other parameter values are as for ( A ).

We explored the sensitivity of the optimal policy to changes in parameter values. Increasing the profitability of cooperation (α) decreases the level of trustworthiness that is required for cooperation to be worthwhile (Fig.  1 B; analytically, the cooperation/foraging alone frontier for \(s \gg 0\) is at \(\left( {1 - p} \right) = 1/\alpha\) ; see Supplementary Sect.  2.2 ). A very high level of social mobility r moves the critical value for exploitation far to the left (i.e. individuals have to be in an even more dire state before they start to exploit; Fig.  1 C). This is because with high social mobility, badly-off individuals can expect that their level of resources will regress towards the mean over time anyway, lessening the need for risky action when faced with a small immediate shortfall.

The optimality of exploitation below the critical level of resources is generally insensitive to increasing the severity of punishment, π (Fig.  1 D), even where the expected value of exploitation is thereby rendered negative. This is because a desperate agent will be below the threshold in the next time step anyway if they forage alone, cooperate, or receive a punishment of any size. They are so badly off that it is relatively unimportant how much worse things get, but important to take any small chance of ‘jumping over’ the threshold. The exploitation boundary is slightly more sensitive to the probability of punishment, γ, though even this sensitivity is modest (Fig.  1 E). When γ is very high, it is optimal for agents very close to the boundary of desperation to take a gamble on cooperating, even where trustworthiness is rather low. Although this is risky, it offers a better chance of getting back above the threshold than exploitation that is almost bound to fail. Nonetheless, it is striking that even where exploitation is almost bound to fail and attracts a heavy penalty, it is still the best option for an individual whose current resource level is desperately low.

We also explored the effect of setting either the probability γ or the severity π of punishment so low that the expected payoff from exploitation is positive. This produces a pattern where exploitation is optimal if an agent’s resources are either desperately low, or comfortably high (see Supplementary Fig. S2 ). Only in the middle—currently above the threshold, but not by far enough that a punishment would not pull them down below it–should agents cooperate or forage alone.

We simulated populations of N  = 500 individuals each following the optimal policy, with the distribution of initial resources s drawn from a distribution with mean μ and standard deviation σ. Populations fall into one of two absorbing equilibria. In the first, the poverty trap (Fig.  2 A), there is no cooperation after the first few time periods. Instead, there is a balance of attempted exploitation and foraging alone, with the proportions of these determined by the initial resource distribution and the values of π and γ. The way this equilibrium develops is as follows: there is a sufficiently high frequency of exploitation in the first round (about 10% of the population or more is required) that subsequent social trust estimates are mostly very low. With trust low, those with the higher resource levels switch to foraging alone, whilst those whose resources are desperately low continue to try to exploit. Since foraging alone produces no surplus, the population mean resources never increases, and both exploiters and lone foragers are stuck where they were.

The two equilibria in simulated populations. ( A ) The poverty trap. There is sufficient exploitation in the first time step ( A1 ) that social trust is low ( A2 ). Consequently, potential cooperators switch to lone foraging, resources never increase ( A3 ), and a subgroup of the population is left below the threshold seeking to exploit. Simulation initialised with μ = 5.5, σ = 4 and all other parameters at their default values. ( B ) The virtuous circle. Exploitation is sufficiently rare from the outset ( B1 ) that trust is high ( B2 ) and individuals switch from lone foraging to cooperation. This drives an increase in resources, eventually lifting almost all individuals above the threshold. Simulation initialised with μ = 5.5, σ = 3 and all other parameters at their default values.

In the second equilibrium, the virtuous circle (Fig.  2 B), the frequency of exploitation is lower at the outset. Individuals whose resources are high form high assessments of social trust, and hence choose cooperation over foraging alone. Since cooperation creates a surplus, the mean level of resources in the population increases. This benefits the few exploiters, both through the upward drift of social mobility, and because they sometimes exploit successfully. This resolves the problem of exploitation, since in so doing they move above the critical value to the point where it is no longer in their interests to exploit, and since they are in such a high-trust population, they then start to cooperate. Thus, over time, trust becomes universally high, resources grow, and cooperation becomes almost universal.

Each of the two equilibria has a basin of attraction in the space of initial population characteristics. The poverty trap is reached if the fraction of individuals whose resource levels fall below the level that triggers exploitation is sufficiently large at any point. With the desperation threshold at s = 0, his fraction is affected by both the mean resources μ, and inequality σ. For a given μ, increasing σ (i.e. greater inequality) makes it more likely that the poverty trap will result, because, by broadening the resource distribution, the tail that protrudes into the desperation zone is necessarily made larger.

The boundaries of the basin of attraction of the poverty trap are also affected by severity of punishment, probability of punishment, and the level of social mobility (Fig.  3 ). If the severity of punishment π is close to zero, there is no disincentive to exploit, and the poverty trap always results. As long as a minimum size of punishment is met, further increases in punishment severity have no benefit in preventing the poverty trap (Fig.  3 A). Indeed, there are circumstances where more severe punishment can make things worse. When the population has a degree of initial inequality that puts it close to the boundary between the two equilibria, very severe punishment (π = 20 or π = 25) pushes it into the poverty trap. This is because any individual that once tries exploitation because they are close to threshold (and is unsuccessful) is pushed so far down in resources by the punishment that they must then continue to exploit forever. Increasing the probability of punishment γ does not have this negative effect (Fig.  3 B). Instead, a very high probability of punishment can forestall the poverty trap at levels of inequality where it would otherwise occur, because it causes some of the worst-off individuals to try cooperating instead, as shown in Fig.  1 E. Finally, very high levels of social mobility r can rescue populations from the poverty trap even at high levels of inequality (Fig.  3 C). This is because of its dramatic effect on the critical value at which individuals start to exploit, as shown in Fig.  1 C.

Equilibrium population states by starting parameters. ( A ) Varying the initial inequality in resources σ and the severity of punishment π, whilst holding constant the probability of punishment γ at 1/3 and social mobility r at 0.1. ( B ) Varying the initial inequality in resources σ and the probability of punishment γ whilst holding the severity of punishment constant at π = 10 and social mobility r at 0.1. ( C ) Varying the initial inequality in resources σ and the level of social mobility r whilst holding constant the probability of punishment γ at 1/3 and the severity of punishment π at 10.

Though the equilibria are self-perpetuating without exogenous forces, the system is highly responsive to shocks. For example, exogenously changing the level of inequality in the population (via imposing a reduction in σ after 16 time steps) produces a phase transition from the poverty trap to the virtuous circle (Supplementary Fig. S3 ). This change is not instantaneous. First, a few individuals cross the threshold and change from exploitation to foraging alone; this produces a consequent change in social trust; which then leads to a mass switch to cooperation, and growth in mean wealth.

Results so far are all based on cooperation occurring in groups of size n  = 5. Reducing n enlarges the basin of attraction of the virtuous circle (Supplementary Sect.  2.5 , Supplementary Fig. S4 ). This is because, for any given population prevalence of exploitation, there is more likely to be at least one exploiter in a group of five than a group of three. Reducing the interaction group size changes the trustworthiness boundary between the region where it is optimal to cooperate and the region where it is better to forage alone. Thus, there are parameter values in our model where populations would succumb to the poverty trap by attempting to mount large cooperation groups, but avoid it by restricting cooperation groups to a smaller size.

In our model, exploiting others can be an individual’s optimal strategy under certain circumstances, namely when their resource levels are very low, and cannot be expected to spontaneously improve. We extend previous models by showing that it can be optimal to exploit even when the punishment for doing so and being caught is large enough to make the expected utility of exploitation negative. Two conditions combine to make this the case. First, exploitation produces a large variance in payoffs: it is costly to exploit and be caught, but there is a chance of securing a large positive payoff. Second, there is a threshold of desperation below which it is extremely costly to fall. It is precisely when at risk of falling below this threshold that exploitation becomes worthwhile: if it succeeds, one hurdles the threshold, and if it fails, one is scarcely worse off than one would have been anyway. In effect, due to the threshold, there is a point where agents have little left to lose, and this makes them risk-preferring. Thus, our model results connect classic economic models of crime 15 , 16 to risk-sensitive foraging theory from behavioural ecology 20 . In the process, it provides a simple answer to the question that has puzzled a number of authors 18 , 19 : why aren’t increases in the severity of punishments as deterrent as simple expected utility considerations imply they ought to be? Our model suggests that, beyond a minimum required level of punishment, not only might increasing severity be ineffective at reducing exploitation. It could under some circumstances make exploitation worse, by pushing punishees into such a low resource state that they have no reasonable option but to continue exploiting. Our findings also have implications for the literature on the evolution of cooperation. This has shown that punishment can be an effective mechanism for stabilising cooperation 28 , 29 , but have not considered that the deterrent effects of punishment may be different for different individuals, due to variation in their states. Our findings could be relevant to understanding why some level of exploitation persists in practice even when punishment is deterrent overall.

Within criminology, our prediction of risky exploitative behaviour when in danger of falling below a threshold of desperation is reminiscent of Merton’s strain theory of deviance 30 , 31 . Under this theory, deviance results when individuals have a goal (remaining constantly above the threshold of participation in society), but the available legitimate means are insufficient to get them there (neither foraging alone nor cooperation has a large enough one-time payoff). They thus turn to risky alternatives, despite the drawbacks of these (see also Ref. 32 for similar arguments). This explanation is not reducible to desperation making individuals discount the future more steeply, which is often invoked as an explanation for criminality 33 . Agents in our model do not face choices between smaller-sooner and larger-later rewards; the payoff for exploitation is immediate, whether successful or unsuccessful. Also note the philosophical differences between our approach and ‘self-control’ styles of explanation 34 . Those approaches see offending as deficient decision-making: it would be in people’s interests not to offend, but some can’t manage it (see Ref. 35 for a critical review). Like economic 15 , 16 and behavioural-ecological 17 theories of crime more generally, ours assumes instead that there are certain situations or states where offending is the best of a bad set of available options.

As well as a large class of circumstances where only individuals in a poor resource state will choose to exploit, we also identify some—where the expected payoff for exploitation is positive—where individuals with both very low and very high resources exploit, whilst those in the middle avoid doing so. Such cases have been anticipated in theories of human risk-sensitivity 21 . These distinguish risk-preference through need (e.g. to get back above the threshold immediately) from risk-preference through ability (e.g. to absorb a punishment with no ill effects), predicting that both can occur under some circumstances 32 . This dual form of risk-taking is best analogised to a situation where punishments take the form of fines: those who are desperate have to run the risk of incurring them, even though they can ill afford it; whilst those who are extremely well off can simply afford to pay them if caught. When we simulate populations of agents all following the optimal strategies identified by the model, population-level characteristics (inequality of resources, level of social mobility) affect the prevalence of exploitation and the level of trust. Specifically, holding constant the average level of resources, greater inequality makes frequent exploitation and low trust a more likely outcome. Thus, we capture the widely-observed associations between inequality, trust and crime levels that were our starting point 2 , 3 , 4 , 5 , 6 . Note that our explanation for the inequality-crime nexus is basically compositional rather than psychosocial. Decisions to offend are based primarily on agents’ own levels of resources; these are just more likely to be desperately low in more unequal populations. Turning these simulation findings into empirical predictions, we would expect the association between inequality and crime rates to be driven by more unequal societies producing worse prospects for people at the bottom end of the resources distribution, who would be the ones who turn to property crime. Inequality effects at the aggregate level should be largely mediated by individual-level poverty. There is evidence compatible with these claims for property crime 2 , 12 , 13 . This is the type of crime most similar to our modelled situation. Non-acquisitive crimes of violence, though related to inequality, do not appear so strongly mediated by individual-level poverty, and may thus require different but related explanations 2 , 36 .

However, the other major result of our population simulations—that more unequal populations are more likely to produce low trust—is not compositional. In our unequal simulated populations, every agent has low trust, not just the ones at the bottom of the resource distribution. This is compatible with empirical evidence: the association between inequality and social trust survives controlling for individual poverty 6 . Thus, our model generates a genuinely ecological effect of inequality on social relationships that fits the available evidence and links it to the psychosocial tradition of explanation 37 . Indeed, the model suggests a reason why psychosocial effects should arise. For agents above the threshold, the optimal decision between cooperation and foraging alone depends on inferences about whether anyone else in the population will exploit. To know that, you have to attend to the behaviour of everyone else, not just your own state. Thus, the model naturally generates a reason for agents to be sensitive to the distribution of others’ states in the population (or at the very least their behaviour), and to condition their social engagement with others on it.

In as much as our model provides a compositional explanation for the inequality-crime relationship, it might seem to imply that high levels of inequality would not lead to high crime as long as the mean wealth of the population was sufficiently high. This is because, with high mean wealth, even those in the bottom tail of the distribution would have sufficient levels of resources to be above the threshold of desperation. However, this implication would only follow if the location of the desperation threshold is considered exogenous and fixed. If, instead, the location of the desperation threshold moves upwards with mean wealth of the population, then more inequality will always produce more acquisitive crime, regardless of the mean level of population wealth. Assuming that the threshold moves in this way is a reasonable move: definitions of poverty for developed countries are expressed in terms of the resources required to live a life seen as acceptable or normal within that society, not an absolute dollar value (see Ref. 36 , pp. 64–6). Moreover, there is clear evidence that people compare themselves to relevant others in assessing the adequacy of their resources 38 . Thus, we would expect inequality to remain important for crime regardless of overall economic growth.

In addition to the results concerning inequality, we found that social mobility should, other things being equal, reduce the prevalence of exploitation, although social mobility has to be very high for the effect to be substantial. The pattern can again be interpreted as consistent with Merton’s strain theory of deviance 31 : very high levels of social mobility provide legitimate routes for those whose state is poor to improve it, thus reducing the zone where deviance is required. Economists have noted that those places within the USA with higher levels of intergenerational social mobility also have lower crime rates 39 , 40 . Their account of the causality in this association is the reverse of ours: the presence of crime, particularly violent crime, inhibits upward mobility 39 . However, it is possible that social mobility and crime are mutually causative.

Like any model, ours simplifies social situations to very bare elements. Interaction groups are drawn randomly at every time step from the whole population. Thus, there are no ongoing personal relationships, no reputation, no social networks, no kinship, no segregation or assortment of sub-groups. The model best captures social groups with frequent new interactions between strangers, which is appropriate since the phenomena under investigated are documented for commercial and industrial societies. A problem in mapping our findings onto empirical reality is that our population simulations generate two discrete equilibria: zero trust, economic stagnation and zero cooperation, or almost perfect trust, unlimited economic growth and zero exploitation. Although we show that the distribution of resources determines which equilibrium is reached, our model as presented here does generate the continuous relationships between inequality, crime, and trust (or indeed inequality and economic growth 41 ) that have been observed in reality. Even the most unequal real society features some social cooperation, and even the most equal features some property crime; the effects of inequality are graded. We make two points to try to bridge the disconnect between the black and white world of the simulations and the shades of grey seen in reality. First, our model does predict a continuous relationship between the level of inequality and the maximum size of cooperating groups. A highly unequal population, containing many individuals with an incentive to exploit, might only be able to sustain collective actions at the level of a few individuals, whereas a more equal population where almost no-one has an incentive to exploit could sustain far larger ones. Second, we appeal to all the richness of real social processes that our model excludes. In unequal countries, although social trust is relatively low, people can draw more heavily on their established social networks and reputational information; more homogenous sub-groups can segregate themselves; people can use defensive security measures, to keep cooperative relationships ongoing and protected; and so forth. Investment in these kinds of measures may vary proportionately with inequality and trust, thus maintaining outcomes intermediate between the stark equilibria of our simulations. Our key findings also depend entirely on accepting the notion that there is a threshold of desperation, a substantial non-linearity in the value of having resources. As we outlined in the Introduction, we believe there are good grounds for exploring the implications of such an assumption. However, that is very different from claiming that the widespread existence of such thresholds has been demonstrated. We hope our findings might generate empirical investigation into both the objective reality and psychological appraisal of such thresholds for people in poverty.

Limitations and simplifications duly noted, our model does have some clear implications. Large population-scale reductions in crime and exploitation should not be expected to follow from increasing the severity of punishments, and these could conceivably be counterproductive. Addressing basic distributional issues that leave large numbers of people in desperate circumstances and without legitimate means to improve them will have a much greater effect. Natural-experimental evidence supports this. The Eastern Cherokee, a Native American group with a high rate of poverty, distributed casino royalties through an unconditional income scheme. Rates of minor offending amongst young people in recipient households decline markedly, with no changes to the judicial regime 42 . Improving the distribution of resources would also be expected to increase social trust, and with it, the quality of human relationships; and this, for everyone, not just those in desperate circumstances.

The model was written in Python and implemented via a Jupyter notebook. For a fuller description of the model, see Supplementary Sect.  1 and Supplementary Table S1 .

Computing optimal policies

We used a stochastic dynamic programming algorithm 25 , 26 . Agents choose among a set of possible actions, defined by (probabilistic) consequences for the agent’s level of resources s . We seek, for every possible value of s and of p the agent might face, and given the values of other parameters, the action that maximises expected fitness. Maximization is achieved through backward induction: we begin with a ‘last time step’ ( T ) where terminal fitness is defined, as an increasing linear function of resource level s . Then in the period T  − 1 we compute for each combination of state variables and action the expected fitness at T , and thus choose for the optimal action for every combination of states. This allows us define expected fitness for every value of the state variables at T  − 1, repeat the maximization for time step T  − 2, and so on iteratively. The desperation threshold is implemented as a fixed fitness penalty ω that is applied whenever the individual’s resources are below the threshold level s  = 0. As the calculation moves backwards away from T , the resulting mapping of state variables to optimal actions converges to a long term optimal policy.

Actions and payoffs

Agents choose among three actions:

Cooperate The agent invests x units of resource and is rewarded α · x with probability 1 −  p ( p is the probability of cooperation being exploited, and 1 −  p is therefore the trustworthiness of the surrounding population), and 0 with probability p . The net payoff is therefore x · ( α  − 1) if there is no exploitation and −  x if there is. We assume that α  > 1 (by default α  = 1 . 2), which means that cooperation is more efficient than foraging alone. For the computation of optimal policies, we treat p as an exogenous variable. In the population simulations, it becomes endogenous.

Exploit An agent joins a cooperating group, but does not invest x, and instead tries to steal their partners’ investments, leading to a reward of β if the exploitation succeeds and a cost π if it fails. The probability of exploitation failing (i.e. being punished) is γ .

Forage alone The agent forages alone, investing x units of resource, receiving x in return, and suffering no risk of exploitation.

Payoffs are also affected by a random perturbation, so the above-mentioned payoffs are just the expected values. A simple form such as the addition of \(\varepsilon \sim N\left( {0, \sigma^{2} } \right)\) would be unsuitable when used in population simulations. As the variance of independent random variables is additive, it would lead to an ever increasing dispersion of resource levels in the population. To avoid this issue, we adopted a perturbation in the form of a first-order autoregressive process that does not change either the mean or the variance of resources in the population 43 :

Here, µ is the current mean resources in the population and σ 2 the population variance. The term \(\left( {1 - r} \right) \in \left[ {0, 1} \right]\) represents the desired correlation between an agent’s current and subsequent resources, which leads to us describing r as the ‘social mobility’ of the population. The perturbation can be seen as a ‘shuffle’. Each agent’s resource level is attracted to µ with a strength depending on r , but this regression to the mean is exactly offset at the population level by the variance added by the perturbation, so that the overall distribution of resources is roughly unchanged. If r  = 1, current resources are not informative about future resources.

The dynamic programming equation

Let I be the set of actions ( cooperate , exploit and alone ), which we shorten as I  = { C,H,A }. For i   ∈   I , we denote as \(\phi_{t}^{i} \left( {s, .} \right)\) the probability density of resources in in time step t if, in time step t  − 1, the resource level is s and the chosen action i . The expressions of these functions were obtained through the law of total probability, conditioning on the possible outcomes of the actions (e.g. success or failure of exploitation and cooperation), and with the Gaussian density of the random variable.

We can now write the dynamic programming equation, which gives the backward recurrence relation to compute the payoff values (and the decisions) at the period t from the ones at the period t  + 1.

Here, \(E_{i}\) is the conditional expectation if action i is played. The optimal action for the time step t is \({\text{argmax}}_{i \in I} E_{i} (f_{t} )\) . The resource variable s was bounded in the interval [− 50, 50], and discretized with 1001 steps of size 0 . 1.

For any given set of parameters (summarised in Table 1 ), we can therefore compute the optimal decision rule. Note that we can distinguish two types of parameters:

‘Structural parameters’, i.e. those defining the ‘rules’ of the game (the payoffs for the actions and the level of social mobility r , for example). In the subsequent simulation phase, these parameters will be fixed for any run of the simulations.

‘Input parameters’, such as p and s . In the simulation phase, these will evolve endogenously.

Optimal policies rapidly stabilize as the computation moves away from T . We report optimal actions at t  = 1 as the globally optimal actions.

Population simulations

We begin each simulation by initializing a population of N  = 500 individuals, whose resource levels are randomly drawn from a Gaussian distribution with a given mean µ and variance σ 2 . At each time step, interaction groups of n  = 5 individuals are formed at random, and re-formed at each time step to avoid effects of assortment. There is no spatial structure in the populations. Each individual always follows the optimal policy for its resources s and its estimate of p (see below). Varying N has no effect as long as N  >  n and 500 is simply chosen for computational convenience.

To deal with the case where several members of the same interaction group choose to exploit, we choose one at random that exploits, and the others are deemed to forage alone (in effect, there is nothing left for them to take). Also, when there is no cooperator in the group, all exploiters are deemed to forage alone.

Rather than providing each individual with perfect knowledge of the trustworthiness of the rest of the population 1 −  p , we allow individuals to form an estimate (their social trust ) from their experience. Social trust is derived in the following way. Each agent observes the decision of a sample of K individuals in the population, counts the number k of exploiters and infers an (unbiased) estimate of the prevalence of exploiters in the population: \(k^{\prime} = \frac{k}{K}N\) (rounded). The size of the sample can be varied to alter the precision with which agents can estimate trustworthiness. Unless otherwise stated we used K  = 50. Since p is the probability that there will be at least one exploiter in an interaction group, it is one minus the probability that there will be zero exploiters. Each agent computes this from their k’ by combinatorics.

An intentional consequence of social trust being estimated through sampling is that there is some population heterogeneity in social trust, and therefore in decisions about which action to take, even for agents with the same resources s . Note also that agents infer trustworthiness not from observing the particular individuals in their current interaction group, but rather, from a cross-section of the entire population. Thus, the estimate is genuinely social trust (the perception that people in society generally do or do not behave well).

Code availability

The Jupyter notebook for running the model is available at: https://github.com/regicid/Deprivation-antisociality . This repository also contains R code and datafiles used to make the figures in the paper.

Tomasello, M. The ultra-social animal. Eur. J. Soc. Psychol. 44 , 187–194 (2014).

Article   Google Scholar  

Kelly, M. Inequality and crime. Rev. Econ. Stat. 82 , 530–569 (2000).

Rufrancos, H. & Power, M. Income inequality and crime: A review and explanation of the time-series evidence. Sociol. Criminol. 1 , 1–9 (2013).

Google Scholar  

Krohn, M. D. Inequality, unemployment and crime: A cross-national analysis. Sociol. Q. 17 , 303–313 (1976).

Barone, G. & Mocetti, S. Inequality and trust: New evidence from panel data. Econ. Inq. 54 , 794–809 (2016).

Kennedy, B. P., Kawachi, I., Prothrow-Stith, D., Lochner, K. & Gupta, V. Social capital, income inequality, and firearm violent crime. Soc. Sci. Med. 47 , 7–17 (1998).

Article   CAS   Google Scholar  

Oishi, S., Kesebir, S. & Diener, E. Income inequality and happiness. Psychol. Sci. 22 , 1095–1100 (2011).

Pickett, K. E. & Wilkinson, R. G. Income inequality and health: A causal review. Soc. Sci. Med. 128 , 316–326 (2015).

Ecob, R. & Davey Smith, G. Income and health: What is the nature of the relationship?. Soc. Sci. Med. 48 , 693–705 (1999).

Nettle, D. Why inequality is bad. In Hanging on to the Edges: Essays on Science, Society and the Academic Lifeg 111–128 (OpenBook Publishers, 2018).

Pridemore, W. A. A methodological addition to the cross-national empirical literature on social structure and homicide: A first test of the poverty-homicide thesis. Criminology 46 , 133–154 (2008).

Machin, S. & Meghir, C. Crime and economic incentives. J. Hum. Resour. 39 , 958–979 (2004).

Raphael, S. & Winter-Ebner, R. Identifying the effect of unemployment on crime. J. Law Econ. 44 , 259–283 (2001).

Wilkinson, R. G. & Pickett, K. E. The Spirit Level: Why Equal Societies Almost Always Do Better . (Allen Lane, 2009).

Becker, G. S. Crime and punishment: An economic approach. J. Polit. Econ. 76 , 169–217 (1968).

Ehrlich, I. Participation in illegitimate activities: A theoretical and empirical investigation. J. Polit. Econ. 81 , 521–565 (1973).

Cohen, L. E. & Machalek, R. A general theory of expropriative crime: An evolutionary ecological approach. Am. J. Sociol. 94 , 465–501 (1988).

Dölling, D., Entorf, H., Hermann, D. & Rupp, T. Is deterrence effective? Results of a meta-analysis of punishment. Eur. J. Crim. Policy Res. 15 , 201–224 (2009).

Nagin, D. S. Deterrence: A review of the evidence by a criminologist for economists. Annu. Rev. Econ. 5 , 83–105 (2013).

Stephens, D. W. The logic of risk-sensitive foraging preferences. Anim. Behav. 29 , 628–629 (1981).

Mishra, S., Barclay, P. & Sparks, A. The relative state model: Integrating need-based and ability-based pathways to risk-taking. Personal. Soc. Psychol. Rev. 21 , 176–198 (2017).

Mishra, S. & Lalumière, M. L. You can’t always get what you want: The motivational effect of need on risk-sensitive decision-making. J. Exp. Soc. Psychol. 46 , 605–611 (2010).

Scott, J. C. The Moral Economy of the Peasant: Rebellion and Subsistence in Southeast Asia . (Yale University Press, London, 1976).

Nettle, D. Tyneside Neighbourhoods: Deprivation, Social Life and Social Behaviour in One British City . (OpenBook Publishers, 2015).

Houston, A. I. & McNamara, J. M. Models of Adaptive Behaviour: An Approach Based on State . (Cambridge University Press, Cambridge, 1999).

Mangel, M. & Clark, C. W. Dynamic Modeling in Behavioral Ecology . (Princeton University Press, Princeton, 1988).

Verducci, S. & Schröer, A. Social Trust. In International Encyclopedia of Civil Society (eds. Anheier, H. K. & Toepler, S.) 1453–1458 (Springer US, 2010). https://doi.org/10.1007/978-0-387-93996-4_68 .

Boyd, R. & Richerson, P. J. Punishment allows the evolution of cooperation (or anything else) in sizable groups. Ethol. Sociobiol. 13 , 171–195 (1992).

García, J. & Traulsen, A. Evolution of coordinated punishment to enforce cooperation from an unbiased strategy space. J. R. Soc. Interface 16 , 20190127 (2019).

Baumer, E. P. & Gustafson, R. Social organization and instrumental crime: Assessing the empirical validity of classic and contemporary anomie theories. Criminology 45 , 617–663 (2007).

Merton, R. K. Social structure and anomie. Am. Sociol. Rev. 3 , 672–682 (1938).

Barclay, P., Mishra, S. & Sparks, A. M. State-dependent risk-taking. Proc. R. Soc. B Biol. Sci. 285 , 20180180 (2018).

Lee, C. A., Derefinko, K. J., Milich, R., Lynam, D. R. & DeWall, C. N. Longitudinal and reciprocal relations between delay discounting and crime. Pers. Individ. Dif. 111 , 193–198 (2017).

Gottfredson, M. R. & Hirshi, T. A General Theory of Crime . (Stanford University Press, Stanford, 1990).

Burt, C. H. Self-control and crime: Beyond Gottfredson & Hirschi’s theory. Annu. Rev. Criminol. 3 , 43–73 (2020).

Daly, M. Killing the Competition: Economic Inequality and Homicide . (Transaction, 2016).

Wilkinson, R. G. & Pickett, K. E. The enemy between us: The psychological and social costs of inequality. Eur. J. Soc. Psychol. 47 , 11–24 (2017).

Payne, B. K., Brown-Iannuzzi, J. L. & Hannay, J. W. Economic inequality increases risk taking. Proc. Natl. Acad. Sci. U. S. A. 114 , 4643–4648 (2017).

Sharkey, P. & Torrats-Espinosa, G. The effect of violent crime on economic mobility. J. Urban Econ. 102 , 22–33 (2017).

Chetty, R., Hendren, N., Kline, P. & Saez, E. Where is the land of opportunity? The geography of intergenerational mobility in the United States. Q. J. Econ. 129 , 1553–1623 (2014).

Alesina, A. & Rodrik, D. Distributive politics and economic growth. Q. J. Econ. 109 , 465–490 (1994).

Akee, K. Q. R., Copeland, W., Keeler, G., Angold, A. & Costello, E. J. Parents’ incomes and childrens’ outcomes: A quasi-experiment. Am. Econ. J. Appl. Econ. 2 , 86–115 (2010).

Bateson, M. & Nettle, D. The telomere lengthening conundrum—it could be biology. Aging Cell 16 , 312–319 (2017).

Download references

Acknowledgements

This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No AdG 666669, COMSTAR). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. The authors thank Melissa Bateson, Juliette Dronne, Ulysse Klatzmann, Daniel Krupp, Kate Pickett, and Rebecca Saxe for their input.

Author information

Authors and affiliations.

Institut Jean Nicod, Département d’études cognitives, ENS, EHESS, PSL Research University, CNRS, Paris, France

Benoît De Courson

Population Health Sciences Institute, Newcastle University, Newcastle, UK

Benoît De Courson & Daniel Nettle

You can also search for this author in PubMed   Google Scholar

Contributions

The project was conceived jointly by B.C. and D.N. The model was developed and programmed by B.C. with feedback from D.N. B.C. wrote the first draft of the paper. The two authors revised and refined it collaboratively.

Corresponding author

Correspondence to Daniel Nettle .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary information., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

De Courson, B., Nettle, D. Why do inequality and deprivation produce high crime and low trust?. Sci Rep 11 , 1937 (2021). https://doi.org/10.1038/s41598-020-80897-8

Download citation

Received : 02 November 2020

Accepted : 23 December 2020

Published : 21 January 2021

DOI : https://doi.org/10.1038/s41598-020-80897-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Leveraging social cognition to promote effective climate change mitigation.

• Mélusine Boon-Falleur
• Aurore Grandin
• Coralie Chevallier

Nature Climate Change (2022)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Poverty as a Harbinger of Crime

• Living reference work entry
• First Online: 01 November 2022
• Cite this living reference work entry

• Oladotun Akanni 8  

Part of the book series: Encyclopedia of the UN Sustainable Development Goals ((ENUNSDG))

32 Accesses

Crime ; Inequality ; Low income ; Population density ; Poverty ; Unemployment

There is no universal definition of poverty because how poverty is defined and measured varies across the world. Perception of poverty or what constitutes poverty depends on the socioeconomic conditions of a particular country. Poverty is complex and multidimensional. Generally, poverty can be ascribed to income deprivation and insufficient resources to meet the minimum and basic necessity of human needs and ensure sustainable livelihoods. Poverty is not just about shortage or paucity of monetary resources but also lack of access to basic social infrastructures like education and healthcare to live a meaningful and decent life. Crime implies any wrongdoing that is liable to punishment in conformity with the law of a country. Simply put, it is the violation of law or an act of lawbreaking. Crimes are committed for various reasons and these include greed, revenge, anger, and envy. Typically,...

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Institutional subscriptions

Adebayo AA (2013) Youths’ unemployment and crime in Nigeria: a nexus and implications for national development. Int J Sociol Anthropol 5(8):350–357

Article   Google Scholar  

Adekoya A, Abdul-Razak N (2016) Effect of crime on poverty in Nigeria. Rom Econ Bus Rev 11(2):30–42

Google Scholar  

Adenike ET (2021) Poverty, unemployment and insecurity challenges in Nigeria. Tanzan Econ Rev 11(1):115–136

Ahad M (2016) Nexus between income inequality, crime, inflation and poverty: new evidence from structural breaks for Pakistan. Int J Econ Empir Res 4(3):133–145

Ali A, Bibi C (2020) Public policies, socio-economic environment and crimes in Pakistan: a time series analysis. Bull Bus Econ 9(1):1–11

Altindag DT (2012) Crime and unemployment: evidence from Europe. Int Rev Law Econ 32(1):145–157

Anser MK, Yousaf Z, Nassani AA, Alotaibi SM, Kabbani A, Zaman K (2020) Dynamic linkages between poverty, inequality, crime, and social expenditures in a panel of 16 countries: two-step GMM estimates. Econ Struct 9(43):1–25

Anwer MA, Nasreen S, Shahzadi A (2015) Social and demographic determinants of crime in Pakistan: a panel data analysis of province Punjab. Int J Econ Empir Res 3(9):440–447

Ayhan F, Bursa N (2019) Unemployment and crime nexus in European Union countries: a panel data analysis. J Adm Sci 17(34):465–484

Baharom AH, Habibullah MS (2008) Is crime cointegrated with income and unemployment?: a panel data analysis on selected European countries. Available at: https://mpra.ub.uni-muenchen.de/11927/1/MPRA_paper_11927.pdf

Berrebi D (2011) Effects of poverty on society, health, children and violence. Available at: https://www.restlessstories.com/poverties/effects-of-poverty

Bharadwaj A (2014) Is poverty the mother of crime? Empirical evidence of the impact of socioeconomic factors on crime in India. Atl Rev Econ 1:1–37

Bhorat H, Lilenstein A, Monnakgotla J, Thornton A, Van Der Zee K (2017) The socio-economic determinants of crime in South Africa: an empirical assessment. Development Policy Research Unit working paper 201704, pp 1–38

Bignon V, Caroli E, Galbiati R (2014) Stealing to survive? Crime and income shocks in 19th century France. IZA – Institute for the Study of Labor, IZA DP no. 8531, pp 1–59

Bjerk D (2009) Thieves, thugs, and neighborhood poverty. IZA – Institute of Labor Economics. IZA DP no. 4470, pp 1–35

Buonanno P (2003) The socioeconomic determinants of crime. A review of the literature. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.504.2832&rep=rep1&type=pdf

Cheteni P, Mah G, Yohane YK (2018) Drug-related crime and poverty in South Africa. Cogent Econ Finance 6(1):1–16. https://doi.org/10.1080/23322039.2018.1534528

Corvalan A, Pazzona M (2019) The ambiguous effects of inequality: the case of crime. Available at: https://sistemas.colmex.mx/Reportes/LACEALAMES/LACEA-LAMES2019_paper_746.pdf

Corvalan A, Pazzona M (n.d.) Does inequality really increase crime? Theory and evidence. Available at: http://www.ecineq.org/ecineq_paris19/papers_EcineqPSE/paper_122.pdf

Dalberis R (2015) Extreme levels of poverty and inequality may lead to equally high levels of social conflict and crime. M.A. thesis, City University of New York (CUNY)

Development Initiatives (2021) Poverty trends: global, regional and national. Available at: https://devinit.org/documents/1100/Poverty_trends_-_global_regional_and_national_-_November_2021.pdf

Dong B, Egger PH, Guo Y (2020) Is poverty the mother of crime? Evidence from homicide rates in China. PLoS One 15(5):1–22

Article   CAS   Google Scholar  

Duque M, McKnight A (2019) Understanding the relationship between inequalities and poverty: mechanisms associated with crime, the legal system and punitive sanctions. CASE paper 215/LIP paper 6. Available at: https://sticerd.lse.ac.uk/dps/case/cp/casepaper215.pdf

ECLAC (2008) Exploring policy linkages between poverty, crime and violence: a look at three Caribbean states. Available at: https://www.cepal.org/sites/default/files/publication/files/5060/S0800466_en.pdf

ECLAC (2022) Social panorama of Latin America, 2021. Available at: https://repositorio.cepal.org/bitstream/handle/11362/47719/1/S2100654_en.pdf

Eurostat (2021) One in five people in the EU at risk of poverty or social exclusion. Available at: https://ec.europa.eu/eurostat/web/products-eurostat-news/-/edn-20211015-1

Fafchamps M, Minten B (2006) Crime, transitory poverty, and isolation: evidence from Madagascar. Econ Dev Cult Chang 54(3):579–603

Gaviria A, Pages C (2002) Patterns of crime victimization in Latin American cities. J Dev Econ 67(1):181–203

Haider A, Ali A (2015) Socio-economic determinants of crimes: a cross-sectional study of Punjab districts. Int J Econ Empir Res 3(11):550–560

Hamzah SNZ, Lau E (2013) The role of social factors in explaining crime. Theor Appl Econ 6(583):99–118

Hooghe M, Vanhoutte B, Hardyns W, Bircan T (2010) Unemployment, inequality, poverty and crime: spatial distribution patterns of criminal acts in Belgium, 2001–06. Br J Criminol 51(1):1–20

Hsieh CC, Pugh MD (1993) Poverty, income inequality, and violent crime: a meta-analysis of recent aggregate data studies. Crim Justice Rev 18(2):182–202

Igbinedion SO, Ebomoyi I (2017) Socio-economic determinants of crime: further evidence from Nigeria. Ann Univ Petroşani Econ 17(1):101–114

Ikejiaku BV (2012) Poverty-conflict nexus: the contentious issue revisited. Eur J Sustain Dev 1(2):127–150

Imran M, Hosen M, Chowdhury MAF (2018) Does poverty lead to crime? Evidence from the United States of America. Int J Soc Econ 45(10):1424–1438

Iyer L, Topalova P (2014) Poverty and crime: evidence from rainfall and trade shocks in India. Harvard Business School working paper, 14-067, pp 1–46

Kelly M (2000) Inequality and crime. Rev Econ Stat 82(4):530–539

Kingston S, Webster C (2015) The most “undeserving” of all? How poverty drives young men to victimisation and crime. Available at: https://eprints.lancs.ac.uk/id/eprint/75784/1/JPSJ_Article_September_2015_final_V5_sent.pdf

Lobonţ Q-R, Nicolescu AC, Moldovan NC, Kuloğlu A (2017) The effect of socioeconomic factors on crime rates in Romania: a macro-level analysis. Econ Res 30(1):91–111

Machin S, Meghir C (2000) Crime and economic incentives. J Hum Resour 39(4):958–979

Maddah M (2013) The effect of unemployment and income inequality on crimes, a time series analysis. Int J Econ Res 4(2):37–42

Malik M, Rabia M, Imran M, Irshad R, Khalid KB (2019) Poverty leads to crime: a case study in GC women university, Sialkot. In: Proc. 15th Islamic countries conference on statistical sciences, Lahore, Pakistan, 21–24 December, 34, pp 107–117

McKeown JE (1949) Poverty, race and crime. J Crim Law Criminol 39(4):480–484

Mehlum H, Miguel E, Torvik R (2006) Poverty and crime in 19th century Germany. Journal of Urban Economics (59)3:370–388.

Ngutu JA (2014) The influence of poverty on crime among the Abanyole of Emuhaya District, Western Kenya. J Humanit Soc Sci 19(4):108–142

Nilsson A (2004) Income inequality and crime: the case of Sweden. IFAU working paper, 2004–6, pp 1–36

Odumosu O (1999) Social costs of poverty: the case of crime in Nigeria. J Soc Dev Afr 14(2):71–85

Olojede D (2015) Ten big ideas. TELL, 23 November, p 43

Omotor DG (2010) Demographic and socio-economic determinants of crimes in Nigeria (a panel data analysis). Available at: http://www.digitalcommons.www.na-businesspress.com/JABE/Jabe111/GodwinWeb.pdf

Osunyikanmi AF (2014) Interrogating the influence of poverty on insecurity in Nigeria. Int J Humanit Soc Sci 4(12):221–227

Owusu G (2016) Introduction: urban crime and poverty nexus. Ghana J Geogr Spec Issue 8(1):1–10

Oyelade AO (2019) Determinants of crime in Nigeria from economic and socioeconomic perspectives: a macro-level analysis. Int J Health Econ Policy 4(1):20–28

Pablo F, Daniel L, Norman L (2000) What causes violent crime? Eur Econ Rev 46:1323–1357

Papaioannou K (2017) Hunger makes a thief of any man: poverty and crime in British colonial Asia. Eur Rev Econ Hist 21:1–28

Patterson EB (1991) Poverty, income inequality, and community crime rates. Criminology 29(4):755–776

Piana V (2006) Poverty. Available at: http://www.economicswebinstitute.org/glossary/poverty.htm

Presidential Committee on the North East Initiative (PCNI) (2016) Rebuilding the North East: the Buhari plan (vol I). Available at: https://www.refworld.org/pdfid/5b42ec184.pdf

Quednau J (2021) How are violent crime rates in U.S. cities affected by poverty? Park Place Econ 28(1):1–24

Rufrancos HG, Power M, Pickett KE, Wilkinson R (2013) Income inequality and crime: a review and explanation of the time-series evidence. Soc Criminol 1(103):1–9. https://doi.org/10.4172/2375-4435.1000103

Shubert C (2003) Female crime and poverty: stolen opportunities. MAIS 701 project, University of Athabasca

Taylor B (2006) Poverty and crime. Available at: http://economics.fundamentalfinance.com/povertycrime

Tesemma G (2017) The nexus between urban crime and poverty: the case of Addis Ababa city administration. M.A. dissertation, Addis Ababa University

Traxler C, Burhop C (2010) Poverty and crime in 19th century Germany: a reassessment. Available at: https://homepage.coll.mpg.de/pdf_dat/2010_35online.pdf

Ucha C (2010) Poverty in Nigeria: some dimensions and contributing factors. Glob Major E-J 1(1):46–56

United Nations (n.d.) Poverty eradication. Available at: https://www.un.org/development/desa/socialperspectiveondevelopment/issues/poverty-eradication.html

Uyang FA, Festus N, Bassey GE (2016) Socio-economic status of youth and involvement in criminal activities in Calabar Metropolis of Cross River State, Nigeria. Int J Humanit Soc Sci Educ 3(1):79–83

Vollaard B, Van Ours JC (2011) Does regulation of built-in security reduce crime? Evidence from a natural experiment. Econ J 121:485–450

Wadsworth TP (2001) Employment, crime, and context: a multi-level analysis of the relationship between work and crime. Ph.D. thesis, University of Washington

Webster C, Kingston S (2014) Anti-poverty strategies for the UK: poverty and crime review. Centre for Applied Social Research (CeASR). Available at: https://eprints.leedsbeckett.ac.uk/id/eprint/849/7/JRF%20Poverty%20and%20Crime%20Review%20June%202016.pdf

Whitworth A (2012) Inequality and crime across England: a multilevel modelling approach. Soc Policy Soc 11(1):27–40

Wikström P-OH, Treiber K (2016) Social disadvantage and crime: a criminological puzzle. Am Behav Sci 60(10):1232–1259

World Bank (2021) Poverty and shared prosperity 2020: reversals of fortune. Available at: https://openknowledge.worldbank.org/bitstream/handle/10986/34496/9781464816024.pdf?sequence=33&isAllowed=y

World Bank, NBS (2018) Conflict and violence in Nigeria: results from the North East, North Central, and South South zones. Available at: https://documents1.worldbank.org/curated/en/111851538025875054/pdf/130198-WP-P160999-PUBLIC-26-9-2018-14-42-49-ConflictViolenceinNigeriaResultsfromNENCSSzonesFinal.pdf

World Vision (2021) Global poverty: facts, FAQs, and how to help. Available at: https://www.worldvision.org/sponsorship-news-stories/global-poverty-facts#multidimensional

WPI Economics (2021) London’s poverty profile 2021: covid-19 and poverty in London. Available at: http://wpieconomics.com/site/wp-content/uploads/2021/04/Londons_Poverty_Profile_2021_-_COVID19__poverty_in_London.pdf

Wu D, Wu Z (2011) Crime, inequality and unemployment in England and Wales. Appl Econ 2011:1–28

Yaacoub S (2017) Poverty, inequality and the social causes of crime: a study between United States and Europe. Int J Sci Res 6(10):629–634

Yang Z (2019) Measurement of childhood poverty in the United States and its enduring influences. J Sociol Soc Welf 46(2):47–73

Zaman K (2018) Crime-poverty nexus: an intellectual survey. Forensic Res Criminol Int J 6(5):327–329

Zaman S, Khan AU (2021) Dynamics of crime rate, income inequality and urbanization across regimes in Pakistan. Indian J Econ Dev 9:1–15

Download references

Author information

Authors and affiliations.

International Development Consultant, Abuja, Nigeria

Oladotun Akanni

You can also search for this author in PubMed   Google Scholar

Editor information

Editors and affiliations.

European School of Sustainability, Hamburg University of Applied Sciences, Hamburg, Hamburg, Germany

Walter Leal Filho

Center for Neuroscience & Cell Biology, University of Coimbra, Coimbra, Portugal

Anabela Marisa Azul

Faculty of Engineering and Architecture, Passo Fundo University Faculty of Engineering and Architecture, Passo Fundo, Brazil

Luciana Brandli

HAW Hamburg, Hamburg, Hamburg, Germany

Amanda Lange Salvia

Istinye University, Istanbul, Turkey

Pinar Gökcin Özuyar

Liverpool Business School, Liverpool John Moores University, Liverpool, UK

Section Editor information

School of Property, Construction and Project Management, RMIT University, Melbourne, VIC, Australia

Usha Iyer-Raniga Ph.D

Rights and permissions

Reprints and permissions

Copyright information

© 2023 Springer Nature Switzerland AG

About this entry

Cite this entry.

Akanni, O. (2023). Poverty as a Harbinger of Crime. In: Leal Filho, W., Azul, A.M., Brandli, L., Lange Salvia, A., Özuyar, P.G., Wall, T. (eds) No Poverty. Encyclopedia of the UN Sustainable Development Goals. Springer, Cham. https://doi.org/10.1007/978-3-319-69625-6_53-1

Download citation

DOI : https://doi.org/10.1007/978-3-319-69625-6_53-1

Received : 17 June 2022

Accepted : 06 July 2022

Published : 01 November 2022

Publisher Name : Springer, Cham

Print ISBN : 978-3-319-69625-6

Online ISBN : 978-3-319-69625-6

eBook Packages : Springer Reference Earth and Environm. Science Reference Module Physical and Materials Science Reference Module Earth and Environmental Sciences

• Publish with us

Policies and ethics

• Find a journal
• Track your research

Income inequality, poverty and crime across nations

Affiliation.

• 1 Department of Sociology, University of Western Ontario.
• PMID: 25251139
• DOI: 10.1111/1468-4446.12083

We examine the relationship between income inequality, poverty, and different types of crime. Our results are consistent with recent research in showing that inequality is unrelated to homicide rates when poverty is controlled. In our multi-level analyses of the International Crime Victimization Survey we find that inequality is unrelated to assault, robbery, burglary, and theft when poverty is controlled. We argue that there are also theoretical reasons to doubt that the level of income inequality of a country affects the likelihood of criminal behaviour.

Keywords: Cross-national criminology; ICVS; homicide; income inequality; poverty.

© London School of Economics and Political Science 2014.

• Crime / economics*
• Crime / statistics & numerical data
• Homicide / economics
• Homicide / statistics & numerical data
• Income / statistics & numerical data*
• Poverty / psychology
• Poverty / statistics & numerical data*
• Theft / economics
• Theft / statistics & numerical data
• Violence / economics
• Violence / statistics & numerical data

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

Is poverty the mother of crime? Evidence from homicide rates in China

Baomin dong.

1 Center for Yellow River Civilization and Sustainable Development, and School of Economics, Henan University, Kaifeng, Henan Province, China

Peter H. Egger

2 Department of Management, Technology and Economics, ETH Zürich, Zürich, Switzerland

3 Institute of Economics and Management, Henan Agricultural University, Zhengzhou, Henan Province, China

Associated Data

The data underlying the results presented in the study are available from PKULAW.COM ( http://www.pkulaw.cn ) and CEIC database ( http://www.ceicdata.com ). Data providers are commercial and access is possible only if users subscribe. Reproducing or publicizing the data to a third party is not permitted.

Income inequality is blamed for being the main driver of violent crime by the majority of the literature. However, earlier work on the topic largely neglects the role of poverty and income levels as opposed to income inequality. The current paper uses all court verdicts for homicide cases in China between 2014 and 2016, as well as various inequality measures calculated from 2005 mini census data together with a host of control variables to shed light on the relationship at the detailed Chinese prefecture-level. The results suggest that it is the poverty and low income level, rather than income inequality, that is positively related to homicide rates. We show that the internal rural-urban migration from more violent localities contributes to the destination cities’ homicide rates. The poverty-homicide association implies that instead of “relative deprivation”, “absolute deprivation” is mainly responsible for violent crime. Poverty is the mother of crime. —Marcus Aurelius (121-180AD), Emperor of the Roman Empire.

Introduction

On June 22, 2017, a mother and her three children died in their home in a tragic fire set deliberately by the family’s migrant worker nanny on the 18th floor of a luxurious high-rise building in Hangzhou, China. The high-profile homicide case in Hangzhou had quickly become a focus of attention for Chinese netizens, with over 200 million views on the Weibo blog within four days. The victim family’s wealth and the nanny’s impoverished and migrant background triggered a nationwide debate on violent crime and tensions over China’s poor-rich divide.

Needless to say, violent crimes such as homicide negatively affect economic development. World Bank [ 1 ] reports that decreasing 10% of a country’s homicides would lift up per capita GDP by 0.7% to 2.9% over the subsequent five years even after controlling for a variety of other determinants. The homicide rate is also an important indicator for rule-of-law [ 2 ]. Homicide is an important dimension of disamenity and can be severely underestimated if the model formulation is incorrect [ 3 , 4 ]. High homicide rates also lower citizens’ satisfaction on democracy, which consequently impair further democratic institutions and economic development [ 5 ]. Policy-makers around the world enact laws to impact public safety, and a controversial one is ‘Stand Your Ground’ on the use of gun violence [ 6 ].

There is a growing literature arguing that income inequality causes crime, beginning with Becker [ 7 ]. The mechanisms seem to be straightforward. For instance, the theory of ‘relative deprivation’ introduced by Merton [ 8 ] suggests that income inequality strengthens the feelings of dispossession and unfairness. The inequality-crime association that empirically identified has been widely accepted. For instance, Messner and Rosenfeld [ 9 ] state that a ‘finding that has emerged with remarkable consistency is that high rates of homicide tend to accompany high levels of inequality’, and solidified in both intra-national and cross-national studies, but it is not without debate.

A fundamental debate centers on whether poverty or inequality drives violent crime. Although many studies show a positive relationship between income inequality and the rate of violent crime, some critical studies using similar data argue that when poverty is added to the regressions, income inequality is not significant anymore and (absolute) poverty explains the crime rate, while (relative) income inequality does not.

Another criticism of the ambiguous effect of inequality on crime in cross country studies is that controlling confounding factors at the country level is inherently difficult. Neumayer [ 10 ] argues that by increasing the sample size of countries in Fajnzylber, Lederman, and Loayza [ 11 ], the Gini coefficient is no longer statistically significant in predicting violent crime. On top of the lack of observability of confounders, income inequality measures for different countries are based on different income concepts and definitions change which places a measurement bias problem, as demonstrated by Atkinson and Brandolini [ 12 ].

In terms of possible reverse causality between income measures (such as various measures of income inequality) and crime rates, the existing literature usually employs an instrumental variable approach but seldom defend the relationship from an intuitive perspective. Nevertheless, it is often argued that the well-offs are able to invest in protection or preventative measures, including installing surveillance system, moving to better communities, and owning a car, etc., thus the reverse causality between property crime and income inequality is also weakened by such argument. In fact, it is documented by Levitt [ 13 ] that households with low income ($25,000 and below) suffer a 60% more chance from burglar than high-income ones ($50,000 and above).

In one study dedicated to the possible reverse causality between crime and inequality, Barenboim [ 14 ] shows that although theoretical model predicts that the rich would spend more on prevention, the empirical evidence from the US indicates that the poor use more expensive transportations to work in areas with the higher violent crime rate. Barenboim [ 14 ] finds a small positive effect of property crime on inequality, but no evidence is shown for such an association between homicide and inequality to exist, to the best of our knowledge. Given that homicide rate is low in absolute magnitude and homicide is largely non-random, it is uneconomic to relocate oneself due to a higher city-level homicide rate.

Another related and commonly believed important factor for violent crimes such as homicide beyond poverty and inequality is immigration. The empirical literature on cross country immigration finds little support of a positive association which runs in contrast to the contrary theory which considers immigration to segregate society and, hence, to stimulate crime. The empirical results seem to support the positive theory, which argues that the fear of deportation, together with the appreciation for a better life often motivate immigrants more than the locals to honor the law. However, the effect of internal migration on local crime rates is understudied by the literature partly due to the lack of data. One exception is Caminha et al. [ 15 ], who find cities with a higher floating population have higher rates of property crimes.

China offers a unique testing ground for the mobility of people and crime for its humongous size of internal migration. According to China Labour Bulletin [ 16 ], there were an estimated 287 million rural migrant workers in China in 2017, making up more than one third of the entire working population. The perception that migrants contribute to most of the crimes may not be mistaken in China. According to a recent popular media analysis [ 17 ], migrants make up over 80% of total violent crimes but the official statistics do not include such information. The Study Group of Xiamen Public Safety Department [ 18 ] also shows that in Chinese cities, the crime rates of immigrants are five to six times higher than those of local hukou residents. However there is a lack of scholarly research on the impact of intra-national migration on violent crimes such as homicide.

The current paper studies the relationship between income measures and homicide rate by using unique data on violent crimes in China, together with mini-census data which give information on the prefecture-level income distribution (allowing to define inequality and poverty rates). We use the “reduced form” model to identify the causality between income measures and homicide rates. Moreover, we use internal migration data between prefectures to control for the violence from migrants. Our results suggest that it is poverty and the income level per se, rather than income inequality, that contributes to the homicide rate. While the rural area poverty level works as a push factor for homicide in a prefecture, boosting crime there, higher average income in urban areas appears to work as a pull factor, attracting crime in the prefecture as a whole.

We believe that the evidence from China in this paper, due to the regionally much more detailed (prefecture-level rather than province-level) account of the dependent and independent variables involved, adds to the ongoing debate on the roles of different dimensions of income and migration for violent crime.

The rest of the paper is organized as follows. Section 2 gives a brief review of the literature. Section 3 describes data and empirical strategy. Section 4 discusses the results. Section 5 concludes the paper.

Literature review

Homicide, as the intentional killing of a person by another, represents the most serious criminal offense in all violent crimes. It is the most widely collected and reported crime in law enforcement and criminal justice statistics. Compared with other types of crime, homicide is least likely to be biased in measurement (e.g., [ 19 – 21 ]). By contrast, e.g., only two-thirds of all burglaries are recorded by the police from victim surveys and police-recorded crime in England and Wales [ 22 ].

A burgeoning literature attempts to explain violent crimes such as homicide by income inequality. The vast majority of the studies show a positive relationship, such as the 34 cross-sectional studies surveyed in Hsieh and Pugh [ 23 ] and the original papers by Lederman et al. [ 24 ], Imrohoroglu et al. [ 25 ], Soares [ 26 ], and Pickett et al. [ 27 ], etc. Notable examples include Blau and Blau [ 28 ] who find such a relationship to exist in the US and Kelly [ 29 ] who concludes that robbery, assault, and the aggregate level of crime are all influenced by income inequality. Messner et al. [ 30 ] find a positive relationship between the Gini coefficient and homicides in the US. Recent evidence for a positive relationship between inequality and the homicide rate is found by Fajnzylber et al. [ 11 , 31 ] for a cross-section of industrialized and developing countries; Poveda [ 32 ] for seven major cities in Colombia; Nadanovsky and Cunha-Cruz [ 33 ] for Latin America; and Demombynes and Ozler [ 34 ] for South Africa.

However, Mathur [ 35 ] finds that the Gini coefficient had an ambiguous effect on crime, and Stack [ 36 ] finds no relationship using data from Interpol for a cross-section of countries. Neumayer [ 10 ] directly questions the results in Fajnzylber et al. [ 11 ] since by increasing the sample size of countries, the Gini is no longer significant in explaining violent crime. Guillaumont and Puech [ 37 ] also do not find a significant impact of income inequality on crime rates. Some other studies, such as Brush [ 38 ] and Chintrakarn and Herzer [ 39 ], even find a negative relationship between income inequality and crime in the US. A common interpretation for this seemingly counterintuitive result is that the larger inequality triggers a larger demand for security devices or services, which leads to a reduction of crime rate. Costantini et al. [ 40 ] report a mixed message in which both inequality and unemployment rate positively impact violent crime where unemployment is often regarded as a proxy of absolute poverty.

Indeed, Pridemore [ 41 ] argues that the positive inequality-homicide association may be a spurious result of model misspecification. In particular, he argues that most cross-national studies of homicide fail to control for poverty, which is the most consistent predictor of area homicide rates in the US empirical literature. In fact, empirical studies implied a poverty-homicide association, e.g., D’Ambrosio and Rodrigues [ 42 ] find a strong spatial correlation between homicide and favelas concentration in Latin America. It is argued that the poverty-homicide association works through disintegrating individuals from society, making them prone to commit to violent crimes [ 43 ].

For the case of China, a number of studies have been carried out to examine the relationship (e.g., [ 44 – 53 ], etc.). Most of them find an inequality-crime association. However, virtually all of the existing papers suffer from problems of small sample size, the use of regionally very aggregated data, the use of aggregated rather than specific types of crime, and relatively imprecise measurement of inequality.

For instance, most of the studies use province-level data, so that the number of observations is small and there is a risk of spatial aggregation bias (income may be relatively equal within but large between meso-regions within provinces, and the relationship to crime within a province may be spurious). Moreover, almost all of the studies took crime data from China Procuratorial Yearbooks in which only aggregate numbers of arrests in each province are reported. This brings multiple difficulties that are hard to overcome. E.g., with aggregate numbers of all arrests, the impact of inequality on specific types of crime may not be studied. The aggregation across different crime types may be of particular concern here, as the relationship between inequality and violent crime depends on the type of crime. For instance, some studies find that income inequality affects homicide rates but not on other types of violent crime such as rape or assault [ 54 – 56 ]. The use of data from China Procuratorial Yearbooks may be a problem since under-reporting [ 57 ] or under-registration [ 58 ] of murder and other violent crimes by the police is a widespread issue.

Regarding the measurement of income inequality, most papers use the urban-rural divide to represent provincial income inequality but its impact on violent crime is hard to justify. E.g., Shi and Wu’s [ 51 ] ‘regional disparity’ is merely the difference between the national average income and provincial income. The most ‘accurate’ estimate of inequality in this line of research, is the provincial Gini coefficient calculated from five quantiles of income grouping recorded in the provincial statistical yearbooks. Similar problems exist for control variables. For instance, the China Procuratorial Yearbooks only give total expenditures of public security agency (police), the procuratorial offices, and the court and judicial agency together, whereas only expenditures on police should directly deter crime. Hence, it is unclear whether counterintuitive results or a lack of significance of the findings should be attributed to the absence of the actual effect, the small sample size, or the aforementioned measurement problems.

Another strand of the literature looks into the relationship between immigration and crime. Despite the widespread perception that immigration is responsible for violent crimes such as homicide, scholarly research finds little empirical evidence for that (see, e.g., in Papadopoulos [ 59 ]). Indeed, numerous studies find that higher concentrations of immigrants are associated with lower crime rates (e.g., Pendergast et al. [ 60 ]). Some studies even establish causality to show that immigration led to lower crime rates. Historically, Sequeira, Nunn and Qian [ 61 ] show that the Age of Mass Migration (1850-1920) had no long-run effects on crime rates in the US. It is also argued that the positive relationship between immigrants and crime in some countries may be due to racial or ethnic discrimination by the police and the judicial system.

However, little is known for the effect of a country’s internal migration on local crime rates. As an exception, Cheng, Liu, and Wang [ 62 ] combine the arrest and prosecution data from 306 prefectures in China with interviews with nine policemen and public procurators from five provinces to show that the ratio of migrants (in the total population) contributes more to the prosecution rate whereas the ratio of home rental over homeownership impacts more on the arrest rate. Thus the implication is that internal migration introduces crime and part of it is carried through the rental-housing channel. In the Chinese language literature, Chen, Li, and Chen [ 63 ] argue that the rise of large-scale internal migration in China is the main reason for the increase in crime rates. However, some other studies, e.g., Tong [ 64 ] finds that migrants are more likely to be victims than criminals due to their low social status and exposure to a complicated environment, etc.

Data and empirical model

The aforementioned data aggregation problem and under-reporting in official statistics make officially published data not well suited for research. However, the movement toward transparency of conviction inflicted by China’s top leader in 2013 resulted in the unconditional disclosure of all court verdicts since January 1, 2014. By decree of the Supreme Court of China, all court verdicts must be made available to public online unconditionally within seven days of judgment, with exceptions of cases of juvenile accused or of national security concerns, beginning by January 1, 2014. Verdicts are collected from PKULaw.cn ( http://www.pkulaw.cn/ ) which is the largest and most inclusive database for legal documents in China. We hand-collected first-instance judgments of all homicide cases from January 1, 2014, to December 31, 2016, from all levels of courts in China, and aggregated them by type to the prefecture level, to form the dependent variable. Thus our data overcomes the incompleteness problem of official statistics, allows a precise measurement of violent crime, and can be measured at the prefecture level rather than province level.

In this study, we use data from PKULaw.cn on all homicides at the level of prefectures that were treated in Chinese courts in 2014 or, alternatively, in 2014-2016. Since the convictions do not reveal any personal information of the victims or the convicted, including second- and/or third-instance information would only bias the data, then we only use the first-stance court convictions and there were in total 8,354 first-stance court convictions in the homicide category between January 1, 2014, and December 31, 2016, published by PKULaw.cn. We normalize the respective numbers by the population size in each prefecture, to obtain the prefecture-level homicide rate. We dub the respective variable Homicide . In general, as the explanatory variables are not available at an annual level for this time span, we will use index i to denote prefectures and use data for 2014 or averages for the years 2014-2016.

We use the following set of explanatory variables to explain Homicide i across prefectures.

Income distribution and poverty

First of all, we use Poverty i , to measure the fraction of the population in the prefecture i , which has an income below the fifth percentile of the overall income distribution of China. Furthermore, we also consider the effect of PovertyUrb i and PovertyRur i in the urban and rural areas of the prefecture i on Homicide i , respectively. Data to compute income percentiles for all of China and for each prefecture underlying the respective poverty variables and variables introduced below are from China’s 2005 mini census of 1% of the population, which is the only dataset with a large-enough sample size to construct China’s prefecture-level income inequality measures. In the 2005 mini census, we use the place of residence (not according to the Hukou type) to define the rural and urban areas. A positive parameter on one of these poverty variables would indicate that an increase in the number of people below the poverty line (in urban or rural areas) would increase the homicide rate in a prefecture.

Moreover, we employ various income-level and income-inequality measures. First of all, we include the log of average income LnAvgInc i , as well as log of average income in urban areas, LnAvgUrbInc i , and rural areas, LnAvgRurInc i , as three separate measures of income in a prefecture. A positive parameter on the average income variables would indicate that an increase in average income (in urban or rural areas) would increase the homicide rate in a prefecture.

Second, we employ three measures of income inequality, all of them also defined for urban and rural areas separately, pertaining to different brackets of the income distribution: (1) the log difference between the 95th and the 5th percentiles, DLnInc 9505 i , DLnUrbInc 9505 i , and DLnRurInc 9505 i ; (2)the log difference between the 90th and the 10th percentiles, DLnInc 9010 i , DLnUrbInc 9010 i , and DLnRurInc 9010 i ; and (3) the log difference between the 75th and the 25th percentiles, DLnInc 7525 i , DLnUrbInc 7525 i , and DLnRurInc 7525 i . While the first measure indicates the income gap in the very tail of the distribution, the last measure captures the interquartile range. The three measures are considered together in order to permit a potentially nonparametric impact of income inequality on Homicide i . A positive parameter on one of these income dispersion variables would indicate that an increase in income dispersion of the respective kind (in urban or rural areas) would increase the homicide rate in a prefecture, even after conditioning on average income and poverty levels per urban or rural area in the prefecture.

We generate the Gini coefficient Gini i for the prefecture i as another measure of inequality between urban and rural area in one additional robustness check. Data for calculating the Gini coefficient is also based on 2005 mini census. A positive coefficient on Gini i would indicate that an increase in income dispersion between urban and rural areas would increase the homicide rate in a prefecture, and such a result is consistent with results using other income inequality measures.

Demography and employment

We employ the following variables and sources to capture demographic factors. LnPopUrb i and LnPopRur i are the log resident population numbers in urban and rural areas of a prefecture. The data source for these variables is CEIC ( www.ceicdata.com ). Moreover, we use the overall population of a prefecture and divide it by the respective area size (in squared kilometers) to obtain population density data, PopDens i . At least for the United States there is evidence that economic crimes are relatively more frequent in more densely populated areas.

Moreover, we employ the ratio of college students in the total population of a prefecture, CollRate i , the unemployment rate, UnempRate i , and the share of manufacturing employment in the total population of a prefecture, IndEmpRate i , all using data from CEIC. Education is also considered in the empirical models since it increases employability, thus decreasing the probability of committing to a crime. Higher levels of unemployment provide direct incentives for economically driven crimes, as justified by Becker [ 7 ], Ehrlich [ 65 ], and Chiu and Madden [ 66 ]. However, we do not have a clear-cut reference point regarding the frequency of homicides in microregions with a higher versus a lower college or unemployment rate, at least not for China.

Policing, jurisdiction, and violent environment

Public security expenditures are hand-collected from all prefecture city governments’ websites on the final settlement of budgetary expenditures for the year 2014. We use the log of prefecture-level public security expenditure and dub it LnPolicing i . Earlier work on crime identified a positive impact of policing on crime numbers, as more policing makes detection and arrests of criminals easier [ 7 ]. Although more policing activities may also deter crimes, the existing literature found no such effect on violent crimes [ 29 ]. Moreover, we employ the ratio of arrests and convictions, ArrConv i . Notice that convictions pertain to a potentially different base of arrested individuals than arrests do. Hence, the respective variable is not necessarily larger than unity. However, in the long run, we expect a larger value of ArrConv i to reflect a less severe threat of the court system for criminals, all else equal. Hence, we expect a positive influence of this variable on Homicide i .

Based on N prefectures, and P provinces, we make twofold use of the N × P matrix of immigration into (cities in) prefectures and the N × 1 vector of intentional injury cases across prefectures from PKULaw.cn. First, using the population distribution across cities, prefectures, and provinces from CEIC data and assuming proportionality about the immigration across prefectures from provinces into cities from the China Migrants Dynamic Survey, we impute an N × N prefecture-to-prefecture immigration matrix where the lines are destinations and the columns are origins. Let us call a typical entry of that matrix m ij , where i is a destination and j an origin. Following customary practice with neighbor weighting in the spatial econometrics literature (see, e.g., Kelejian and Prucha [ 67 ]), we normalize m ii = 0 for all prefectures i . Moreover, let us refer to ι j as a typical entry of the vector of prefecture-level intentional injury cases. Then, we generate the variable NeighborViolence i based on ∑ j = 1 N m i j ι j ∑ j = 1 N m i j . Hence, NeighborViolence i is the immigration-weighted number of intentional injury cases in other prefectures than a given one. Similarly, we define ViolentImmig i as ∑ j = 1 N ι j m i j ∑ j = 1 N ι j - ι i , which is the intentional-injury-weighted number of immigrants from other prefectures. For controlling the net inflow poor migrants’ effect on homicide, we generate another variable ImmigPoor i based on the immigration matrix and the number of residents living in the minimum subsistence allowances in 2014 from statistical yearbook of each prefecture. We define ImmigPoor i as ∑ j = 1 N m i j ω j - ∑ i = 1 N m j i ω i to resprensent the net inflow of poor migrants to prefecture j , where ω j and ω i are the proportion of the residents living in the minimum subsistence allowances in prefecture j and prefecture i .

A positive parameter on NeighborViolence i would indicate that homicides are higher in prefectures which are located in the vicinity of prefectures with high rates of intentional injuries. Hence, this variable reflects one source of potential cross-border spillover effects of violent crime. A positive parameter on ViolentImmig i would indicate that homicides are higher in prefectures which are more important destinations of migrants from prefectures with high rates of intentional injuries. Hence, this variable reflects another source of the potential cross-border spillover effect of violent crime. A positive parameter on ImmigPoor i would indicate the homicides are higher in prefectures which had more net inflow of migrants. By adding this variable, we are able to identify whether poverty imported via inflows of migrants contributed to local violent crime.

Region fixed effects

Apart from these continuous variables, we also include binary indicator variables for seven geographical divisions often used to aggregate Chinese provinces. These regions are: Eastern China (Anhui, Fujian, Jiangsu, Jiangxi, Shandong, Shanghai, Zhejiang); Southern China (Guangdong, Guangxi, Hainan); Northern China (Beijing, Hebei, Inner Mongolia, Shanxi, Tianjin); Central China (Henan, Hubei, Hunan); Northeastern China (Heilongjiang, Jilin, Liaoning); Southwestern China (Chongqing, Guizhou, Sichuan, Tibet, Yunnan); Northwestern China (Gansu, Ningxia, Shaanxi, Xinjiang).

Descriptive statistics

We provide a list of the variables just introduced together with a short definition and descriptive statistics in Table 1 . For each variable used, we provide the number of prefectures this variable would be available for ( Obs .), the average value the variable takes on ( Mean ), as well as the corresponding standard deviation (Std.dev.), the minimum ( Min .), and the maximum ( Max .).

We will suppress a lengthy discussion of the descriptive statistics for the sake of brevity but will highlight a few facts. The degrees of poverty vary widely across different prefectures in China, as well as average income and income inequality also show regional differences. The apparent contrast of poverty and income inequality also can be seen when we consider urban and rural areas separately. First of all, average urban income is higher than average rural income, as can be seen from the average values of the variables LnAvgUrbInc and LnAvgRurInc . Moreover, the average inequality is higher in urban than in rural areas for any inter-quantile range considered. To see this, compare the average values of DLnUrbInc9505 , DLnUrbInc9010 , and DLnUrbInc7525 with those of DLnRurInc9505 , DLnRurInc9010 , and DLnRurInc7525 , respectively. Observe also that the average rural poverty is higher than its urban counterpart (see the statistics for PovertyUrb vs. PovertyRur ).

Urban areas tend to be smaller than rural ones according to the averages of LnPopUrb and LnPopRur , but the largest urban area is larger than the largest rural one across the prefectures (see the maxima of the same variables).

Finally, consistent with our earlier discussion, the ratio of arrests to convictions of crimes is not bounded by unity from below. To see this, consider the maximum value of ArrConv in Table 1 .

Empirical model

Generally, the dependent variable and covariates may be endogenous due to simultaneity, and it shall be treated with care. In the case of the lack of convincing instrumental variables, a reduced form regression using lagged variables is suggested by the literature. Following Angrist and Pischke [ 68 ], we use a reduced-form model to identify the causality. Suppose we have such structure model with the true regression coefficient we are interested in:

where indices i for prefectures, t for year, vector X i , t is an endogenous variable, and W i , t is the disturbance. Consider using the poverty and inequality measure in 2005 as instrumental variables, which is the lag of our endogenous income measures. Because the lag of income measures affects the current term directly, and cannot play any role in the current homicide rate, the true structure model in the first stage is then:

Thus the reduced form is the regression of the dependent variable on the instrument X i , t −1 given by:

where U i , t is the disturbance. Assuming E ( X i , t −1 W i , t ) = 0, that is, the lag of the poverty and inequality affects homicide rates only through affecting themselves in 2014, and E ( X i , t −1 X i , t )≠0. It is credible to assert that the income measures in the past can be proper IVs to affect homicide rates in the current.

To reduce the notational burden, we eliminate the subscript of year t and include all the aforementioned prefecture-level variables including our IV into the vector Y i . Using a for seven geographical divisions (in which prefectures are nested) and denote region fixed effects by μ a , we can write the empirical model we estimate as:

where G (·) indicates that we estimate a generalized-linear exponential-family model to accommodate the non-negativity of the dependent variable Homicide i , and U i is the respective disturbance term which obeys E ( U i ) = 1 and U i ≥ 0 and which we treat as potentially heteroskedastic. As Homicide i ∈ [0; 1) is defined as a rate that is bounded from below, and from above, we follow Papke and Wooldridge [ 69 ] and use a fractional-response framework. Typically, G (·) rests on a logistic function G ( z ) = exp( z )/(1 + exp( z )), which maps z to the(0, 1) interval. Then we maximize the binomial log likelihood with individual contribution given by:

One point noteworthy in the reduced form regressions is that the effect might be biased because the income level (or income inequality) in 2005 may have had an effect on 2014 income levels. Then the coefficient ( β ) here contains the whole effect which includes direct effect ( γ ) and indirect effect ( α ) of income inequality in 2014 on crime rates in 2014. From the above model, we may estimate coefficients with upward bias, however, the reduced form does not change the direction of the causality. At the same time, using the income measures in 2005 as an IV alleviates the problem of endogeneity due to simultaneity.

We put regression results in five tables. Tables ​ Tables2 2 and ​ and3 3 present the results at the whole prefecture level. Table 4 presents the results in which the relevant income measures are calculated at prefectural urban and rural levels. Table 5 reports robustness checks where inequality measures are substituted by Gini coefficients.

We report the p-value of LR test and χ 2 value ( p-value in parentheses) of H-L test in the last two rows. Robustness standard errors in parentheses;

* p < 0.1,

** p < 0.05,

*** p < 0.01.

In all regressions we control the region fixed effects. We choose the binomial model structure and the logit link function by using the link test which based on Pregibon’s work [ 70 ] to test the rationality of the family model structure for the fractional-response framework. We also perform the Hosmer-Lemmeshow (H-L) test to evaluate the goodness of fit of our model. When the p-value of the test is greater than 0.05, then we cannot reject the null hypothesis that the observed and expected proportions are the same across all quantiles, thus the model fits well. Based on the Liklihood Ratio (LR) test and H-L test, all the estimations we run have a high goodness of fit.

Table 2 has seven numbered columns, which differ in terms of the included income measures as well as other demographic factors. Columns (1)-(6) suggest that a higher income difference between the 95th and the 5th percentiles induces lower homicide rates, but higher income differences between the 90th and the 10th percentiles ( DLnInc9010 ), and the 75th and the 25th percentiles ( DLnInc7525 ) induce higher homicide rates. Only the latter two income inequality measures provide consistent results with the existing literature even though all these three measures are not statistically significant. When three measures are considered together to show the nonparametric impact of income dispersion on homicides in column (7), only the coefficient of DLnInc9010 is statistically significant and positive. This indicates that an increase in income dispersion would increase the homicide rate at 90th and 10th percentile income inequality. However as we shall demonstrate later, this inequality-income association may be spurious once absolute poverty is introduced in the regressions.

As for the control variables, we find that a higher arrest-to-conviction rate ( ArrConv ) reduces prefecture-level homicide rates, and higher intentional-injury-weighted number of immigrants ( ViolentImmig ) induces higher homicide rates. These estimated coefficients appear to be important at conventional levels of significance across columns (2)-(7). Other covariates’ effects on homicide rate conform intuition and some are statistically significant, e.g., total rural population, population density, and industrial sector employment. Thus, the results in Table 2 roughly replicate the positive inequality-crime association found in many existing studies, albeit the relationship is not statistically significant in general.

Based on the results in Table 2 , we add poverty level and average income in the model and the results are shown in Table 3 . The results in columns (1)-(3) suggest a clear adverse effect of a greater degree of poverty ( Poverty ) and higher income level ( LnAvgInc ) of a prefecture appears to raise homicides there. However, columns (4)-(6) imply that the same poverty and income-level factors are insignificant but positive after controlling for the dispersion of income and other demographic covariates. Inequality in the tails ( DLnInc9505 and DLnInc9010 ) still show the significant and even larger values compared to Table 2 . Likewise, only the coefficient of DLnInc9010 is significantly positive which means an increase in income dispersion at 90th and 10th percentile income inequality would increase the homicide rate significantly. Therefore, our main results in Table 3 attest that all the poverty level, income-level, and income-dispersion factors show an adverse effect in homicide rate in prefecture-level. Moreover, the poverty-level and income-level appear to have more important and larger impacts than income inequality.

As for the control variables in Table 3 , we find that the expenditures on policing ( LnPolicing ) reduces homicides significantly across columns (1)-(3) but changes to be positive and insignificant when adding inequality measures and other demographic covariates. This may arise from the other factors such as population, unemployment and income inequality play larger roles in homicide than public security expenditure. Similarly, the net inflow of poor migrants ( LnImmigPoor ) becomes positive on the homicide rate after adding other poverty and income inequality variables. The small and statistically insignificant effect of LnImmigPoor does not support the claim that the local homicide rate is largely determined by the poverty of net inflow migrants. However, this is not to say that violent crime is not related to internal migration, indeed, our variable, ViolentImmig which captures the spatially weighted influence of violence propensity due to the net inflow of migrants, positively contributed to the local homicide rate. Thus these results indicate that it is the violence per se that imported through migrants, rather than imported poverty, that escalates violent crime in a prefecture.

In order to conclusively decide on an appropriate factor closely related to homicides, we investigate further by comparing the difference between rural and urban areas of prefectures. Table 4 gives us more complicated results based on Table 3 by dividing key explanatory variables into urban and rural areas.

Relative to the insights from Table 3 , we find that the same poverty, income-level, and income-dispersion factors matter when considered in urban and rural areas, respectively. A higher degree of poverty in the rural areas ( PovertyRur ) of a prefecture on homicides there, and also a higher income level in the urban areas ( LnAvgUrbInc ) of a prefecture appears to raise homicides there. However, columns(4)-(6) imply that a greater dispersion of incomes does not—neither in the urban nor the rural part of a prefecture—increase homicide rates in the prefecture. Hence, these results attest to the fact that poverty in the rural parts to be a push factor and higher income levels in the urban parts to be a pull factor of homicide rates. However, they do not attest to an important adverse effect of income inequality on prefecture-level homicides.

Furthermore, the effects of control variables reported in Table 4 are consistent with those in Table 3 , implying the robustness of our estimation. The latter findings suggest that a negative impact of college education ( CollRate ) on homicide rates, and a positive impact of a higher degree of unemployment rates ( UnempRate ) on homicides. Both urban and rural population decrease homicide rates as shown by the negative and statistically significant coefficients of LnPopUrb and LnPopRur , which is in contrast to the empirical patterns observed in Latin American countries [ 71 , 72 ].

We also do a robustness test by replacing the income inequality measures into the Gini coefficient ( Gini ) in Table 5 . All columns in Table 5 show that a higher Gini coefficient (foremost in urban but even in rural areas) of a prefecture, regardless of overall, rural, or urban, reduces homicide rates, negating the inequality-crime nexus. At the same time, a higher poverty level, particularly in rural areas (both in terms of statistical significance and magnitude), drive up homicide rate. For the pull factor, a higher average income-level (particularly in urban areas) increases homicide rates. These findings are manifest the robustness of our benchmark results.

In the remainder of the discussion, we focus on the effects of log urban average income ( LnAvgUrbInc ) and rural poverty rates ( PovertyRur ) on homicide rates, using nonlinear (fractional-logit) predictions of the homicide rate in column (7) of Table 4 .

In Fig 1 , we plot predicted homicide rates when evaluating all variables except PovertyRur and LnAvgUrbInc at their sample average. For generating the figure, we use a 21 × 21 grid of equally-spaced cells in the maximum-to-minimum range of PovertyRur and LnAvgUrbInc in the data. Consistent with the fractional-logit form, this obtains a three-dimensional nonlinear (S-shaped) function. The steepest ascent of this function would be in the neighborhood, where Homicide ≈0.5 with respect to LnAvgUrbInc , and 0.66 with respect to PovertyRur . However, the data are situated around a much lower mean. Accordingly, the three-dimensional nonlinear function is ascending in PovertyRur-LnAvgUrbInc -space throughout the sample. Hence, the steepest ascent in either dimension is at the maximum value of both PovertyRur and LnAvgUrbInc at the point where Homicide is highest in Fig 1 .

In Figs ​ Figs2 2 and ​ and3, 3 , we plot the marginal effects of these two variables for each prefecture in the data by way of a map. In each figure, we increase the respective variable by one standard deviation from the value a prefecture had in the outset. In this exercise, we do not use the means of all the other control variables, but we account for their prefecture-specific values. Fig 2 is devoted to the response of homicide rates to an increase in rural poverty, and Fig 3 to an increase in average urban income. The two figures suggest that there is a large degree of heterogeneity in the responses, and the effects tend to be largest in the northern prefectures and lower in the central and southern parts of China. In general, the effect of PovertyRur is much larger (and also its level of statistical significance is higher, which is not obvious from the maps) than that of LnAvgUrbInc .

Increase in Rural Poverty.

Increase in Urban Average Income.

Conclusions

There is a burgeoning literature on the relationship between income inequality and violent crimes. The majority of this line of literature attributes the widened inequality as a driven force of higher violent crimes. However, the relationship may be spurious once some poverty measure is added, as argued by a handful of studies. The current paper provides support to the latter argument using court convictions on homicide throughout China, mini-census data on incomes, and migration survey data, all at the prefecture level.

We find that it is the poverty level in rural areas and the average income level in urban areas, rather than income inequality, that contribute to the local rate of incidence of violent crimes in China. This finding is robust against different measures of inequality and different empirical formulations. Apart from poverty in rural areas, the ‘transferred’ violence by internal immigrants pushes the crime rate.

The current paper therefore contributes to the ongoing literature by providing prefecture-level empirical evidence from China. Our study calls for a reexamination of the robustness of the results in earlier work by adding poverty measures in the regressions. The inequality-homicide association appears to likely disappear and the poverty-homicide association to hold, at least in China. Our purpose is not to argue that inequality is unimportant, but when it comes to violent crime, “absolute deprivation” appears to matter more than “relative deprivation”.

Funding Statement

This study was supported by the Center of Yellow River Civilization and Sustainable Development; China Scholarship Council (grant number: 201808410386).The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability

• Search Menu
• Advance articles
• Editor's Choice
• Author Guidelines
• Submission Site
• Open Access
• Why Publish with EREH?
• About European Review of Economic History
• About the European Historical Economics Society
• Editorial Board
• Advertising and Corporate Services
• Journals Career Network
• Self-Archiving Policy
• Dispatch Dates
• Contact the European Historical Economics Society
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

1. introduction, 2. historical context, 4. empirical strategy, 5. main empirical results, 6. conclusion, supplementary material, statistical yearbooks and government reports, “hunger makes a thief of any man”: poverty and crime in british colonial asia.

• Article contents
• Figures & tables
• Supplementary Data

Kostadis J. Papaioannou, “Hunger makes a thief of any man”: Poverty and crime in British colonial Asia, European Review of Economic History , Volume 21, Issue 1, February 2017, Pages 1–28, https://doi.org/10.1093/ereh/hew019

• Permissions Icon Permissions

This study uses rainfall variation as an instrumental variable for rice production to estimate the impact of poverty on different types of crime across British colonies in South and South East Asia (1910-–1940). Using original primary sources retrieved from annual administrative and statistical reports, it provides some of the first evidence in a historical setting on the causal relationship between poverty and crime. Extreme rainfall, both droughts and floods, lead to a large increase in property crimes (such as robbery, petty theft, and cattle raiding), but not to an increase in interpersonal violent crimes (such as murder, homicides, and assault). In line with a growing body of literature on the climate-economy nexus, this study offers evidence that loss of agricultural income is one of the main causal channels leading to property crime. Additional historical information on food shortages, poverty, and crime is used to explore the connection in greater detail.

Climate change and its potential threatening impacts have spurred a growing body of literature examining how extreme weather conditions influence economic performance and human behaviour. This literature suggests that deviations from average rainfall and temperature levels increase the likelihood of intergroup conflict ( Fjelde and von Uexkull 2012 ; Hsiang et al . 2013 ), inter-communal conflict ( Bai and Kung 2011 ; Papaioannou 2016 ), onset of civil war ( Blattman and Miguel 2010 ; Miguel et al . 2004 ), property crime ( Blakeslee and Fishman 2014 ; Iyer and Topalova 2014 ; Mehlum et al . 2006 ), civil unrest and disobedience ( Christian and Fenske 2015 ) and even complete institutional breakdowns ( Brückner and Ciccone 2011 ). However, some of this evidence has been contested on both theoretical ( Buhaug et al . 2014 ) and methodological grounds ( Klomp and Bulte 2013 ; Sarsons 2015 ), leaving the debate far from settled.

The societal impact of weather variability seems to be stronger and less ambiguous in developing regions, where the majority of cultivated crops are rain-fed and an insignificant share of cultivated areas is equipped with irrigation and artificial water drainage systems. 1 Unfortunately, we still do not fully understand the underlying mechanisms driving the climate-to-economy relationship ( Burke et al . 2015 ). The most commonly hypothesized channel is that of falling incomes and, by extension, poverty ( Dell et al . 2014 ; Hsiang et al . 2013 ; Miguel et al . 2004 ). In a predominantly agrarian society—being primarily a rain-fed economy—economic prosperity is intimately tied to agricultural output. Extreme weather conditions—resulting in drought or flood—are associated with poor harvests and complete crop failures. Consequently, loss of a year's harvest, besides bringing about near-famine conditions, can easily push farmers into extreme poverty.

Poverty has long been a question of great interest within a wide range of fields. Multiple scholarly disciplines, including economics, political science, history, and anthropology, have observed and documented that poverty and crime go hand in hand. The literature distinguishes between absolute poverty (i.e., lack of minimal material necessities for survival) and relative poverty (i.e., extreme income inequality). A great deal of previous research has demonstrated that absolute poverty is associated with higher property crime rates ( Iyer and Topalova 2014 ; Mehlum et al . 2006 ; Patterson 1991 ), while relative poverty has been linked with the surge of aggression and violent crime ( Blau and Blau 1982 ; Fajnzylber et al . 2002 ; Kelly 2000 ). Throughout this study, the term “poverty” will be used in its broadest definition to encompass a wide range of conditions such as abrupt food shortages, starvation, hunger, subsistence crises, and near-famine conditions.

In simple economic theory of crime, originally introduced by Becker (1968) , individuals are more likely to become involved in criminal activity when they experience a negative income shock. This reasoning is framed in terms of an opportunity cost model ; as income levels decline as a result of unfavourable conditions, engaging in crime becomes more opportune relative to participating in more “peaceful” economic activities ( Grossman 1991 ; Seter 2016 ). While the theoretical foundations of poverty and crime have been well-established, the empirical basis for such an argumentation is considered speculative at best ( Burke et al . 2015 ; Dell et al . 2014 ). One plausible explanation for this omission is the endogenous relationship between poverty and crime: deteriorating economic conditions may favour criminal activity, since more people are likely to engage in crime as an alternative source of income, whilst higher levels of crime may undermine economic stability, investment, and productivity. In other words, does poverty generate crime—or does crime lead to more poverty? Or does some third factor, for example state's institutional capacity or certain food policy reform, affect both simultaneously?

Previous studies have been unable to resolve the key econometric identification issues and have been potentially subject to bias due to reverse causality and omitted variables, both of which distort simple ordinary least squares (OLS) estimates either downward or upward. For instance, OLS estimates of the effect may be biased downwards if colonial governments are more likely to invest in food relief programs in districts that experience high crime rates. These investments will underestimate the poverty effect. On the other hand, OLS estimates of poverty on crime would be biased upwards if, for example, high crime rates bring about higher poverty. Likewise, an upward bias may result from third factors, such as the occurrences of political and economic crises, that tend to increase both crime and poverty simultaneously. In our findings we show that simple OLS estimates are, indeed, biased downward and underestimate the impact of poverty on crime to a substantial degree. Although this article's aim is to apply this approach in economic history, it may also be extended to more present day developing countries.

British imperialism in South and South East Asia, c. 1914.

Our key hypothesis is two-fold: 1.  If weather shocks lead to crime through subsistence crises, then these shocks should primarily affect the kinds of crime that alleviate loss of income. We argue that this may very well be the case in both rural and urban areas but for different reasons. In rural areas, farmers are directly affected by the deficient harvest and resort to illicit activities, whereas in urban areas, waged labourers have to cope with food-price spikes, since they are much more dependent on the market for their daily calories ( absolute poverty). 2. If weather-induced harvest failures are causing sharp increases in income inequality (relative poverty), for instance because some farmers or merchants benefit from exceptional market power during a period of food scarcity and food-price hikes, we would expect more violent uprisings and grievances against people who were making money by exploiting the needs and misery of others. In other words, on the one hand, income losses caused by weather shocks should primarily increase petty theft and cattle raiding, and, on the other hand, perceptions of “injustice”, “exploitation”, or “abuse” of miserable circumstances may induce a rise in violent crime such as homicides, murders, and assaults.

We investigate this possibility by distinguishing between these two broad crime categories; i.e., property and violent crimes. We find that a one standard deviation decrease in annual rice production increased property crime by 21.2 percent. This effect is considerably higher in magnitude to accumulated evidence from other studies reviewed by Hsiang et al . (2013) . There are three possible reasons why the effect is larger than that which the literature predicts. The first reason has to do with the institutional context in which this study is embedded and the limited attention the colonial governments paid to local food production. 4 Despite the fact that by the early twentieth century most colonial governments were taking a more activist approach to promoting widespread economic development in the territories they controlled ( Booth 2012 ), it remains highly refutable the extent to which they managed to convert growing national output into improved living standards for the vast majority of local population. Colonial government expenditures on transport infrastructure, food relief programs, and irrigation/drainage systems remained considerably low in per capita terms. 5 Conservative fiscal policies and the enduring food versus cash crop substitution dilemma among colonial officials impeded local padi-rice cultivation and substantially increased reliance on imported foodstuffs ( Elson 1997 ; Lim 1976 ).

A second reason why the effect was so large has to do with the substantially low living standards prevailing among rural communities at the time. While exports may have boomed and government revenues expanded, nutritional intakes for the mass of the population did not improve, and as a result mortality rates were high ( Booth 2012 ). That the effect between rainfall shocks and property crime rates declined considerably throughout the twentieth century, suggests that modern economic growth and its ensuing technological improvements in agriculture have enhanced the ability of vulnerable rural societies to withstand climate-induced calamities. Lastly, in seeking to explain the high magnitude of the effect, it should be reasoned that we are dealing with a non-industrial part of the world, where the vast majority of the total income was derived from agricultural practises, and where wage labour opportunities were limited, and wage rates low. ( Bray 1994 ; Drabble 1973 ; Elson 1997 ; Farmer 1977 ).

In line with our second hypothesis, it is found that the effect of poverty on violent crimes was insignificant and nearly zero, suggesting extreme income inequality was not (as) crucial in inciting crime. Additionally, it is shown that a standard deviation decrease in rice yields increased begging and vagrancy by 13.8 percent, suggesting that rice production was a key determinant of poverty during this period.

This study yields three contributions. First , we find strong evidence that both drought and excessive rainfall cause substantial increases in property crime. The results are robust to using alternative econometric specifications, to using lagged dependent and independent variables, to cross-sectional spillovers, and finally, to clustering standard errors at the country level, the year level as well as two-way clustered at both the country and the year level. These results are in accord with recent studies indicating that property crimes are more likely to increase in years of depressed incomes than violent crimes ( Blakeslee and Fishman 2014 ; Iyer and Topalova 2014 ; Mehlum et al . 2006 ; Miguel 2005 ), as well as with simple economic theory of crime ( Becker 1968 ; Grossman 1991 ; Bourguignon 2000 ).

Second , a causal relationship is established between poverty and property crime in an agrarian historical setting, and the channel of causality is identified by using rainfall as a source of exogenous variation in food production. While it is intuitively plausible that the rainfall instruments are exogenous, we have to evaluate whether they satisfy the exclusion restriction: rainfall shocks should affect property crime only through reduced agricultural production. One potential concern may be a direct causal link (without any changes to real income) between high temperatures and aggressive and/or violent behaviour, as several psychological and empirical studies have documented ( Anderson 1989 ; Anderson et al . 2000 ; Blakeslee and Fishman 2014 ; Iyer and Topalova 2014 ; Ranson 2014 ). Our results confirm this direct extra-economic channel, since high temperature shocks are associated with more violent crimes (coeff. = 0.043, s.e. = 0.018). However, temperature shocks yield no association with property crimes nor with adverse effects on food production. This serves as an important validation of the empirical strategy and highlights the importance of looking beyond aggregate crime measures in this climate-crime literature, since they may shadow heterogeneous patterns across crime categories.

Another potential violation of the exclusion restriction is the possibility that rainfall deviation may have a direct effect on crime; if heavy rains, due to flooded roads for instance, reduce criminals’ likelihood of being detained by the police or hamper police's capacity to report crimes. Therefore, if such channels are present, IV estimates could misattribute the direct effects of rainfall to crime. Note that such channels are not serious threats to our estimation strategy, since excessive rainfall is empirically associated with significantly more (not less) crime in the reduced form regressions. Thus to the extent that a bias exists, our estimates would be lower bounds on the true impact of poverty on property crime.

A critical assumption underlying the use of rainfall as an IV is that rainfall shocks affect property crime only through their impact on agricultural income . The most prominent critique put forward for such an assumption is that rainfall shocks could also potentially operate through non-agricultural urban channels (manufacturing and non-agricultural wages). However, these channels are unlikely to be important in the agrarian and largely non-industrial context we study, where rain-fed agriculture dominates aggregate production and a relatively small amount of urban labourers existed ( Bruton 1992 ; Drabble 1973 ; Parmer 1960 ; Tregonning 1965 ). We acknowledge that our identification strategy may be inappropriate for other more developed regions of the world, where rainfall shocks are not sufficiently related to depressing agricultural income, for instance, due to substantial investment in irrigation and drainage infrastructure ( Sarsons 2015 ). Nevertheless, this strategy is likely to be of interest to both economic historians and policy makers, since it is highly viable for poor agrarian societies, such as pre-industrial Europe or contemporary less developed countries.

Third , we make use of additional historical information on food shortages and crime as reported in the colonial sources. This is a substantial contribution since most studies in this field are solely based on econometric correlations and make no attempt to contextualize their findings using qualitative evidence. Using a systematic approach, we were not only able to confirm the empirical findings of this study but also to find support in favour of the theoretical foundations of the opportunity cost model. We observe that as foodstuffs become scarcer and distress levels rise, property crimes, and vagrancy inflate substantially—while violent crimes remain unaffected.

Our work also relates to the emerging literature on the effects of weather shocks on crime and conflict. Interestingly, the vast majority of the studies published on the climate-economy nexus has been mostly restricted to limited comparisons in the post-1960 period, mainly due to data availability (see Dell et al . 2014 ; Hsiang et al . 2013 ). Only very recently scholars have begun expanding the temporal scope in the pre-1960 period (for example, Bai and Kung (2011) and Jia (2014) for premodern China; Anderson et al . (2015) and Bignon et al . (2016) for premodern Europe; Papaioannou (2016) and Papaioannou and DeHaas (2017) for colonial Africa). This study thus expands both the geographical as well as temporal scope of studies on the climate-economy nexus.

The remainder of the paper proceeds as follows. Section 2 reviews the historical context and the colonial reports. Section 3 describes the data sources and the construction of variables used in the analysis, and Section 4 lays out the empirical strategy. Section 5 presents our empirical findings and Section 6 concludes.

In thinking about the possible mechanisms underpinning the relationship between food shortages, poverty, and crime in British colonial Asia, it is important to consider some of the underlying agricultural, economic, and political conditions. We begin by providing a brief historical background, including a discussion of the importance of rice production and various aspects of the British colonial rule. We then proceed by reviewing the colonial reports to offer further contextualization, to shed light on the underlying mechanisms, and to validate the theoretical foundations of the opportunity cost model .

2.1 Historical background

During their long history as agrarian societies, South and South East Asian states were very vulnerable to unfavourable climatological conditions ( Bray 1994 ; Hill 2012 ). Peasant agriculture failed to rise above subsistence level during the period of British rule ( Bruton 1992 ; Tregonning 1965 ). The crops continued to be adversely affected by natural enemies and there were limited agricultural advances to increase yields as the peasant's technology had changed only slightly over the years and the growth of agricultural production was further impeded by the rapid expansion of rubber cultivation in the early twentieth century ( Bray 1994 ; Drabble 1973 ; Elson 1997 ). Even though padi rice was the traditional staple crop in this part of the world, the quantities produced were sometimes described as being inadequate to meet the wants of the people who grow it and several states were, to various degrees, dependent on imported agricultural produce ( Butcher 1979 ; Farmer 1977 ). However, padi rice production formed the principal occupation of the peasant, and was the chief source of general prosperity ( Elson 1997 ; Lim 1976 ).

Many historians have observed that deep in the peasant's ethos was the understanding that padi rice was the foundation of life and its cultivation the proper and most honoured sphere of toil. To be a cultivator very nearly meant to be a padi cultivator ( Bray 1994 ; Farmer 1977 ). As Hill (2012 , p. 59) puts it “ the Malaya states were predominantly agricultural and rice-growing was so prevalent that had led to the virtual exclusion of any other food crop” . For most Malaya states the proportion of rice growers to total economically active persons was above 80 percent. Nevertheless, the Malays did not attempt double-cropping as wet land rice required so long a period to reach maturity, that there would have been a deficiency of water for a second crop. In cases where a second, light crop was attempted, “ it often scarcely repaid the trouble of cultivation; so poor was its yield ” ( Hill 2012 , p. 111). A similar story can be deduced from Mills (1964) and Bruton (1992) for Ceylon, and from Tregonning (1965) for North Borneo. Tregonning (1965 , p. 93) points out that “ although the area devoted to rice cultivation was greater than that given to any other crop and its culture was the chief industry of the native people, there was never enough rice produced for the territory to be self-sufficient, and large quantities were always imported ”.

The significance of padi-rice cultivation for the overall well-being of the native population was also documented in the colonial sources. As we will see in more detail in the next section, the agricultural commissioners were required to report whether food supply was sufficient and well maintained to cover local needs. If rice production was scarce, they were requested to provide plausible explanations. Time and again, they pointed to climatological conditions in explaining the deficiency or failure of the annual padi-rice harvest and its consequential distress among smallholders.

The basic function of British colonization in the Asian states covered by this study, whatever its form and origin, was to establish and maintain the conditions under which the dynamic forces of trade could flourish ( Butcher 1979 ; Lange 2004 ; Mills 1964 ; Parmer 1960 ; Wade 1998 ). Although the British rule was intended to be indirect, direct rule was the practice and it increased in scope and effectiveness as the years passed. The new structure was purely British in conception and operation. British interests dominated the vast majority of commercial activities as the British owned most of the plantations, the estate factories that processed the rubber, tea, and coconuts, the import-export trade, and other service activities. In areas with favourable soils and climates, cash crop export economies (such as coffee, tea, rubber etc.) were encouraged and promoted as they provided the colonial authorities with much needed revenue from customs duties and other forms of indirect taxation ( Booth 1999 ; Drabble 1973 ).

2.2 Qualitative evidence: climate, poverty, and crime

The sources used are the annual reports filed by the colonial administration. The British colonizers set up an extensive system of administration, where elaborate administrative accounts were kept. These accounts make regular notice of weather-induced agricultural failure resulting in higher levels of social distress, and in more extreme cases, subsistence crises, and near-famine conditions. In practice, each department filed reports to provide information on and explanations of various incidences along with what was considered to be their causes. The goal of this section is not to systematically record all incidences of food shortages, subsistence crises, and crime, but rather to give a detailed contextual overview of the reasons, mechanisms, and channels appearing to be most relevant.

2.2.1 Rainfall extremes: flood and drought

a drought at the wrong season or a flood will cause great loss to the country and to the padi planters. It was at one time roughly estimated that the present year's padi crop would, owing to drought, be some 25.000.000 gantangs of padi short of the previous year's gantang. As a result it would involve a prospective loss of $2.500.000 to padi planters. (Kedah Annual Report , 1925).

In 1934, the same commissioner reported a severe crop failure, owing this time to excessive rainfall: “ s erious flooding was experienced during November and early December, doing extensive damage in several districts, and it is believed that the harvest will be less abundant than that of the previous year ”. His speculation came true by the end of that year when “ the total yield was just about 33 million gantangs, showing a decrease of 17 million gantangs as compared with 1933 ” (Kedah, Annual Report , 1934).

Likewise, in 1924, the Kelantan colonial official recorded both types of unfavourable weather conditions within the State and remarked that “ the high variability of yields has been connected chiefly with droughts and floods. As one reads back through the annual reports, one constantly comes across reports of disastrous droughts. At other times it is an early flood which has drowned the padi before it has become sufficiently established ”. The rice returns reveal that “ the total production of rice has fluctuated rather more violently than the area planted, and sometimes in an opposite direction. Thus in the 1924 season the 148.000 acres yielded only 55,359 tons, whereas last year the 139,000 acres yielded 74,008 tons ” — a decrease in production of ca 25 percent. The reason for such a decrease was the “ serious flooding during the early part of the season made it impossible for peasants to cultivate their land ”. Thus, overall, our qualitative evidence supports a U-shaped parameterization of the link between weather and agricultural incomes.

2.2.2 Food scarcity and price spikes

The annual reports are very extensive and meticulous in the way they describe local agricultural conditions. These reports frequently mention the adverse consequences of weather conditions in inducing food shortages, food-price hikes, and resultant social distress. Shortages of rice and other foodstuffs and subsequent increases in their prices were greatly felt by the people. In years of food scarcity, local padi-rice was sold for high prices, which were beyond the means of the poor.

that there was a large increase in the number of offences against property. A number of causes are given for this increase…among them the most probable seem to be economic due to unfavourable weather conditions. The last reason applies more particularly to the surrounding plain, where thieves and house-breakers, when detected, are usually found to be young padi planters (Trengganu, Annual Report, 1914).

Similarly, the agricultural commissioner of Central province in Ceylon recounted in 1927 that “ the padi-planters are going through a year of almost unprecedented misfortune ” and that “ the abnormal rainfall in January did considerable damage. That was a truly phenomenal rainfall. Paddy cultivation has been a failure—one of the lowest crops on record, and it was followed by the usual food shortage. This is also shown by the fact that burglars take away foodstuffs, which formerly were left behind. Owing to the shortage of paddy crops the price of paddy has risen dramatically. Deaths from starvation are occasionally reported ”. Indeed, the rice crop was a poor one and totalled 300.000 gantangs as compared to 1.200.000 the year before—a decrease of ca 75 percent (Ceylon, Annual Report, 1927). The following two cases were selected to offer a comprehensive impression of the far-reaching impact of rice shortages on the vast majority of households, across numerous aspects of ordinary life, regardless of whether they were net rice importers or exporters. One could hypothesize that the effect might be stronger in rice-importing districts, as the available food supplies did not meet local demand. Thus, when the already insufficient padi-rice harvests fail, economic distress would be inflated. By contrast, one could argue that the effect might be stronger in rice exporting districts, as the annual rural income of households would be more exposed and sensitive to weather fluctuations. Thus, a failed rice harvest would have a relatively more acute impact on rice exporting districts than in rice-importing ones. While these dynamics are more systematically addressed in the econometric part of the paper, we illustrate the broad social impact of rice shortages in cases where rice was exported and imported.

For instance, in Selangor (a rice-importing state), when the rains failed and available incomes shrank, “ the health of the people deteriorated and fell considerably below the normal standard. This was to a large extent due to the acute scarcity of food. Enteric cases soared in almost all towns ” (F.M.S. Medical Report, 1919). Similarly, in 1936 Kedah (a rice exporting district), the adverse effects of food scarcity were felt on the general welfare of the people. Infant mortality rose by 11 percent and average daily attendance at government schools declined from 13.425 in 1935 to 9.912 in 1936 (approx. 27 percent decrease). “ The decrease was due to the scarcity of rice and the poverty of the people. Children were in many cases half starved, and their parents could not provide them with clothes ”. (Kedah, Education Report , 1936).

2.2.3 Poverty and property crime

there is a great temptation to them [padi-growers] to commit crime in order to live; and reports received indicate that there has been an increase of cattle-stealing and theft of foodstuffs. Thefts of foodstuffs have not been common in the past, and the increase in this type of crime and in cattle-stealing is due to real poverty and difficulty in obtaining food (Ceylon, Annual Report, 1931).

the year was probably the worst ever experienced in the State. The difficulties to be faced were great. Crop after crop was destroyed and by June the shortage of food was becoming very serious. Theft was rife and increased as food became short (North Borneo Native Affairs, 1919).

The number of offences against property—cattle lifting, burglary, and petty thefts—has shown a most unsatisfactory tendency. This is doubtless due to the scarcity of food. I have heard it said that the way in which private gardens are rifled is a real deterrent to enterprise in their cultivation. This may be an exaggeration, but the evil is rampant, and in some cases is caused by real starvation.

it was impossible to obtain padi in the villages…The harvest was late, the crops were entirely ruined…rice crops were everywhere poor and in many places destroyed by the phenomenal drought…The presence of property crime is undoubtedly due in no small measure to a shortage of rice (F.M.S., Judicial Report, 1932).

Few would dispute that livestock breeding broadens the opportunities to store wealth, mediate risks, and raise land productivity in pre-industrial societies. However, livestock was also seen as an object of looting, since by stealing few cattle in times of hardship the perpetrator gains either income by marketing the cattle and exchanging it for other goods, or gains calories by simply consuming it. This is exemplified best in the report undertaken by the colonial officer of Trengganu state in 1931, who related the unfavourable weather conditions of that year with a considerable rise in arrests, stressing that many parts of the central plain were “ so infested with thieves that poultry and cattle could not be kept and was stolen by night ”.

In many instances the colonial officials associated food shortages directly with cattle-raiding noting that “ there was an increase of cattle theft, perhaps due to the shortage of food. Most cattle stolen are slaughtered ” or that “ cattle stealing was rife in North Kedah during the first nine months of the year owing to the unfavourable weather conditions of the country in certain localities ” .

Simple theoretical considerations suggest that income shocks should have a larger impact on property crimes as compared to violent ones. Our qualitative material confirms that logic, as the colonial officials reported sizable differences between the amount of property and violent crimes within the same year. We observe many instances where increases in property crime did not yield concurrent increases in violent crimes. According to the 1919 Annual Report, Kedah had “ experienced in succession the two driest years since rainfalls were first recorded in 1906 and as is anticipated, the padi crop reaped at the beginning of 1919 was poor ”. Theft returns that year more than doubled; from 403 up to 812 cases—an increase of more than 100 percent, whereas the violent offences against the person declined substantially. The Commissioner reports that “ there is a large amount of petty thieving but there are remarkably few crimes of violence ” .

The evidence presented thus far supports the idea that scarcity of food and loss of income had led to substantially more property crime. However, unlike years of extreme weather fluctuations, there have also been seasons with exceptionally good yields. The impact of “feast” seasons on reducing crime levels was noted by the colonial administrators as well. In years where the precipitation patterns were smooth, the rice-harvest was bountiful, and as a result crime rates plummeted. According to the 1930 Agriculture report of Kelantan: “ the crop was harvested under ideal weather conditions and proved to be one of the best secured for a number of years. A surplus over the requirements of the indigenous population was obtained ”. According to the Police report of that same year “ There has been little serious crime, owing to prosperous favourable weather conditions. ”[…] “ as regards criminal litigation, increased prosperity has led to diminution of crime ” (Kelantan, Police Report , 1930).

Summary statistics: district by year

Notes : Author's calculations. See main text.

3.1 Crime data

Among the crimes included are theft, cattle-raiding, assault, and homicide. Of the crimes included in the data, we combine individual crime categories into two broad categories: theft and cattle raiding as property crimes , and assault and homicide as violent crimes . Results are presented for both individual and aggregate crimes. Additionally, data are obtained for vagrancy for each state they were available. All variables exhibit a high year-to-year variation. Table A1 presents the pairwise correlation matrix among the dependent variables. In thinking about issues of differential crime reporting over time, it is important to note that we have no evidence to believe that weather shocks would affect the incentive of crime victims to report crime or the incentive of colonial officials to record more or less criminal incidences. Even though it is not implausible that local authorities would in fact increase the level of under-reporting at precisely those times when crime is rife—i.e., after an adverse weather shock—we argue that this tendency would, if anything, downward our results. 6 The summary statistics for the types of crime are presented in panel ( a ) of table 1 .

3.2 Weather shocks

Rainfall shocks and property crime.

Since there are many ways to parameterize rainfall, we also transformed our main explanatory variable, and defined a “negative rainfall shock” as a dummy which takes the value of 1 when annual rainfall in a district i is one standard deviation below the long-run mean of panel i , and a “positive rainfall shock” as a dummy being one standard deviation above the panel's mean. We also use an alternative measure of rainfall taken from the Matsuura and Willmott (2009) database. The data have 0.5° latitude by 0.5° longitude grids. Values for standardized grid rainfall deviation have a mean of 0.00, a standard deviation of 1, and range from −2.87 to 3.38. Although this measure gives nearly identical results, it reduces  1. our sample size considerably (Singapore, Perlis and a few districts in North Borneo had to be dropped due to lack of observations) and 2. the accuracy of the observations as the data compiled in this dataset is mostly based on extrapolation. Lastly, we construct a variable of temperature deviation to account for the extra-economic direct impact of extreme heat on crime. The summary statistics of the weather conditions are presented in panel ( b ) of table 1 .

3.3 Rice production

Rainfall shocks and food production.

3.4 Controls

While omitted variables should not be of great concern, a number of additional time-varying controls have been included to address potential bias stemming from unobserved factors. We control for differences in 1. state capacity by constructing a measure of road density following Herbst (2014) , 2. policing capacity by constructing a measure of police staff density ( Papaioannou 2016 ), and 3. demographic pressures by constructing a measure of population density. Lastly, we control for the interaction of a few spatial characteristics of districts with a linear time trend to take into account their heterogeneous impacts over time. We construct a battery of district-specific effects to control for the possibility that some districts would react differently over time. It should be reminded here, that we are dealing with a timespan of about three decades, and we, therefore, expect some unobservable characteristics at the district level to change over time. Thus, by including the interactions of district dummies and a linear time trend, we allow the estimates to take into account widening differences across regions and districts during the long-time horizon of this study. For instance, we expect that the colonial authorities in the early 1940s (as compared to the early 1910s) to have extended their capacity to broadcast power and to have become more effective in opposing crime.In addition, it could be that a spike in property crimes in district i , may have urged the colonial authorities to invest more in those crime-stricken places and, as a result, in the following years the capacity of police in inhibiting future crime would have been increased. In either case, not accounting for such systematic tendencies in the data might have yielded inconsistent estimates.

We first establish that rainfall shocks significantly affect crime rates (reduced form specifications) and then proceed by focusing on identifying the causal channel linking poverty and crime (first-stage and IV-2SLS results).

The impact of rainfall shocks on crime (reduced form)

Notes: *Significant at 10 percent, **5 percent, ***1 percent. Sample period: 1910–1939. OLS-FE. The dependent variables are the logarithm of each crime variable expressed as 100,000 of the population. Reported in parentheses are standard errors clustered at the district level. Controls include population density, road density, and police per capita. District-specific effects indicate the interaction of each District dummy × Trend.

In achieving our second goal and identifying the causal effect of poverty on crime, we present the OLS first-stage relationship results between rainfall and food production, and then perform a IV-2SLS estimation using rainfall as an instrumental variable for food production. While few scholars have put the use of rainfall variation as instrument for income under scrutiny ( Sarsons 2015 ), by pointing to alternative non-agricultural channels (e.g., urban wages or direct psychological effects) through which rainfall shocks may increase crime, a large body of literature argues that the use of this kind of instrumentation approach in rainfed agrarian settings is highly suitable ( Burke et al . 2015 ; Dell et al . 2014 ; Mehlum et al 2006 ; Miguel 2005 ; Miguel et al . 2004 ). In either case, we conduct several robustness checks to address potential violations of the exclusion restriction (Section 5.4 ).

The first part of this Section ( 5.1 ) presents the results on the impact of weather shocks on the various types of crime (reduced form specifications). Section 5.2 presents the results of the instrumental variable approach (first-stage and two-stage IV-2SLS). Section 5.3 presents a set of heterogeneous effects for different subsamples. Section 5.4 presents the results for a set of robustness checks and Section 5.5 refutes potential violations of the exclusion restriction.

5.1 The impact of weather shocks on crime (reduced form)

Magnitude of coefficients.

The impact of lagged and lead rainfall shocks on crime

Notes : *Significant at 10 percent, **5 percent, ***1 percent. Sample period: 1910–1939. OLS-FE. The dependent variables are the logarithm of each crime variable expressed as 100.000 of the population. Reported in parentheses are standard errors clustered at the district level. Controls include population density, road density, and police per capita. District-specific effects indicate the interaction of each District dummy × Trend.

Curvilinear impact of rainfall shocks on property crime

Notes : *Significant at 10%, ** 5%, ***1%. Sample period: 1910–1939. OLS-FE. Reported in parentheses are standard errors clustered at the district level. The dependent variable is the logarithm of the property crime variable expressed as 100.000 of the population. The estimated coefficients can be interpreted as percentage changes. District-specific effects indicate the interaction of each District dummy × Trend.

We also test for the symmetricity of the effect by including the “positive rainfall shock” and “negative rainfall shock” variables into the analysis. A standard deviation increase in rainfall increased property crimes by 17.7 percent and similarly, a standard deviation decrease in rainfall increased property crimes by 8.3 percent. This result is in line with previous findings by Papaioannou (2016) for Nigeria and Papaioannou and DeHaas (2017) for colonial British Africa. A possible explanation is that in years of excessive rainfall farmers would lose their entire harvest in a relatively shorter time, whereas in years of drought farmers could hope for late rains. In the former case, the certainty of a failed harvest more rapidly reduces the opportunity cost of crime. Another possible explanation is that in tropical regions with relatively abundant rainfall regimes, excessive precipitation causes flooding and acute surface run-offs which, in turn, hurt harvests considerably more.

5.2 Poverty and property crime: an instrumental variable approach

Rainfall shocks and food production (first stage)

Notes : *Significant at 10%, **5%, ***1%. Sample period: 1910–1939. OLS-FE. Reported in parentheses are standard errors clustered at the district level. The dependent variable is the standardized annual rice production (z-score). Controls include population density, road density and police per capita. District-specific effects indicate the interaction of each District dummy × Trend. The F-statistic of our preferred specification (column 3) is 18.3.

Baseline IV-2SLS results: poverty and crime

Notes : *Significant at 10%, **5%, ***1%. Sample period: 1910–1939. OLS-FE. Reported in parentheses are robust standard errors clustered at the district level. Controls include population density, road density and police per capita. District-specific effects indicate the interaction of each District dummy × Trend. The instrumental variable is rainfall deviation at year t . Vagrancy statistics were not available for the whole sample.

We next estimate the impact of loss of income on vagrancy arrests. Under the British colonial rule, begging was illegal and destitute people ended up in police reports. Even though vagrancy can hardly be characterized as crime, it could serve as a suitable proxy for dire poverty. We expect vagrancy to yield a similar robust correlation as property crime. Indeed, the results show that the relationship between food production and vagrancy is negative and highly significant, which suggests that depressed incomes were a major determinant of poverty. One standard deviation decrease in rice production increases the amount of arrested vagrants by 13.8 percent (regression 6). Table A6 reports the IV-2SLS results on violent crime. All the estimated coefficients are nearly zero.

5.3 Heterogeneous effects

Heterogeneous effects: compliers

Column 4 limits the sample to districts that showed a lower percentage of padi-rice cultivation than the median district, and Column 5 limits the sample to districts that showed a higher percentage of padi-rice than the median. Both coefficients are statistically different from zero, and they are not statistically distinguishable from each other (p-value = 0.183). Despite the slightly higher coefficient of districts with a lower percentage of padi-rice production than the median district (coeff. = −0.181, s.e. = 0.029), the results suggest that the vast majority of households suffer a loss in income when food production is low, regardless of the total volume of per capita rice produced in that particular district. 9 This finding is in line with Iyer and Topalova (2014) , who find that rainfall shocks in India decrease average consumption across the full range of the income distribution, impacting all segments and production classes of society.

Columns 6 and 7 divide the sample by whether the district was exporting considerable volumes of cash crops (mainly rubber, coconut, tea, and cocoa) than the median district. The correlation between food production and property crime is statistically significant in both samples, and it is substantially larger in the less commercial agricultural sample (coeff. = −0.312, s.e. = 0.096) than in the more commercialized sample (coeff. = −0.157, s.e. = 0.088). Moving from the more commercialized districts to the less commercialized ones, the relationship becomes more pronounced and property crime increases by almost half. We argue that the widening gap seen in the estimated coefficients is due to the lack of economic diversification. It seems likely that agricultural commercialization and crop diversification acted as an insurance mechanism to local households by generating an alternative source of income. This result is in line with that of Papaioannou and DeHaas (2017) for colonial British Africa and Burgess and Donaldson (2010) for colonial India; both effectively arguing that crop diversification and openness to trade mitigated the adverse effects of weather shocks.

Next, we hypothesize that districts possessing less infrastructural density experienced higher transportation barriers and costs and were more difficulty reached by potential food relief programs. Columns 8 and 9 divide the sample by whether the district had less barriers to market access than the median district. We proxy market access with infrastructural density. We expect districts with relatively more dense road network to be less susceptible to shocks, since high road density facilitates inter-regional and international trade. The correlation between food production and property crime is statistically significant in both samples but is substantially larger in the districts with higher infrastructural density. The two coefficients are statistically distinguishable from each other (p-value = 0.038). The greater responsiveness of property crimes to food shocks in more isolated districts is a reflection of the scarcity of alternative income opportunities from trade. Columns 10 and 11 divide the sample by whether a district had a higher level of public expenditure than the median district. The reasonable assumption is that districts with a relatively higher budget could intervene and invest in years of agricultural loss; hence, the poverty-crime effect would be attenuated. Nevertheless, we find this not to be the case. The difference between the two samples is insignificant (p-value = 0.311).

Lastly, Columns 12 and 13 tackle potential concerns related to the likelihood of some districts receiving more rainfall shocks than the median district. To achieve that, we have transformed our rainfall data following the coefficient of variation (CV) formula, also known as relative standard deviation. This is a standardized measure of rainfall dispersion which is expressed as a percentage. It is defined as the ratio of the standard deviation σ i to the mean μ of each district i . The values for CV rainfall range from 0.0919 to 0.2574. The sample is split by whether the district is more likely to face a rainfall shock than the median district. Our results show that while both are statistically different from zero, they are not statistically distinguishable from each other (p-value = 0.167).

5.4 Robustness checks

We now check the robustness of the preferred IV-2SLS estimates as reported in table 6 . First, Table A2 shows that replacing rainfall deviation obtained from meteorological stations with an alternative measure of rainfall, based on the Matsuura and Wilmott (2009) world rainfall database (0.5 × 0.5 grid), gives nearly identical results. Second, account is taken of widening differences across countries as well as heterogeneity during the 30-year horizon of this study and Table A3 , presents the IV-2SLS results of standard errors clustered at the country level, year level as well as two-way clustered at both the country and year level.

Third, we examine the sensitivity of the main estimates to the use of alternative instrumental specifications. Column 1 of Table A4 reports estimates using rainfall in the lagged year t − 1 and two years earlier t − 2 as instrumental variables. Similarly, as an identification check, we estimate a “false experiment” specification in which leads of rainfall deviation in year t + 1 and t + 2 are included as instrumental variables, and find that the coefficient estimate is indeed near zero (column 2). As an additional falsification test, we re-estimate our main IV-2SLS results by using temperature shocks as an instrumental variable (column 3). These checks provide additional validation to our empirical strategy.

Table A5 reports the IV-2SLS estimates for each individual category of crime, and table A6 for violent crimes. In results not reported, we obtain statistically identical results, if we use standardized beta coefficients ( z -scores) for transforming the main dependent variables. Thus, the IV-2SLS results are not specific to the choice of functional form. Lastly, to ensure that our results are not driven by spatial spillovers, since rainfall patterns could be spatially correlated, we control for spatial and serial correlation using methods suggested by Hoechle (2007) . The results remain largely unchanged.

5.5 Potential violations of the exclusion restriction

While it is intuitively plausible that the rainfall instruments are exogenous, we have to evaluate whether they satisfy the exclusion restriction—i.e., weather shocks should affect property crime only through falling agricultural income. We acknowledge the possibility that economic channels (either direct or indirect ones) other than annual rice production may affect crime in the aftermath of adverse rainfall shocks. One possible violation of the exclusion restriction may occur in the case when rainfall shocks directly impact on crime; in an extreme rainfall scenario, flooded roads for instance, may reduce criminals’ likelihood of stealing due to transportation difficulties or may hamper police capacity to report crimes. If such channels are present, IV estimates could misattribute the direct effects of rainfall to poverty. Note though that such alternative explanations do not pose a serious threat to the estimation, since excessive rainfall is associated with more (not less) crime in the reduced form regression (coeff. = +0.172 in table 4 column 7). Thus to the extent that a bias exists, our estimates would be lower bounds of the true impact of poverty on property crime.

Another possible concern is that the colonial states may have intervened by investing more in places with extreme poverty. If extreme poverty was declining, and property crimes were to a large extent driven by poverty, one might expect the impact of food production on crime to decrease over time. To test for such a concern, we include interaction terms between food production and a time trend, which we instrument with interactions between rainfall shocks and a time trend. However, we do not find support for the claim that the effect of poverty on property crime attenuated during the study period (results not reported).

Another possible channel is psychological, as rainfall may affect people's moods by making them more or less inclined to commit a crime. A clear candidate here is high temperature shocks which have been found to cause elevated aggression ( Anderson 2001 ) and violent crimes ( Ranson 2014 ). We find that temperature shocks are not positively or negatively associated with rice yields ( Table A7 , column 1) nor with property crimes (column 2). However, consistent with the relevant literature ( Anderson et al . 2000 ; Blakeslee and Fishman 2014 ; Iyer and Topalova 2014 ), we find that temperature shocks are associated with 4.3 percent more violent crimes (column 3).

This article suggests that income shocks, and by extension poverty, are a key underlying cause of property crime in British colonial Asia. We estimate the causal effect of reduced rice production on crime using rainfall variation as an instrumental variable for rice production, and find that the effect of abrupt income shocks on property crime is considerably large. A one standard deviation decrease in annual rice production increases property crime by 21.2 percent. This effect is considerably higher in magnitude to accumulated evidence from other studies reviewed by Hsiang et al . (2013) . One explanation for such a large effect may arise from the fact that we are dealing with a highly agrarian/non-industrial part of the world, where the vast majority of the total income is derived from agricultural practises such as livestock herding and (food and cash crop) farming, and where urban labour was limited. Another possible explanation has to do with the institutional context in which this study is embedded and the limited attention the colonial governments paid to local food production. Lastly, another explanation that has been put forward has to do with the substantially low living standards prevailing among rural communities at the time, where nutritional intakes for the mass of the population did not improve, and as a result mortality rates were high.

With the use of rainfall as an instrumental variable for padi-rice production, this article also addresses a methodological challenge; i.e., endogeneity and reverse causality, since the effect between poverty and crime is larger than simple OLS estimates would suggest (8.5 percent), highlighting the importance of using instrumental variable methods. Additionally, we show that a one standard deviation decrease in rice production increases begging and vagrancy by 14.1 percent. This finding suggests that rice production was a key determinant of poverty during this period.

Although we find no effect between income shocks and violent crime, our results confirm a direct extra-economic channel between high temperature and violent behaviour. A one standard deviation increase in temperature is associated with 4.3 percent increase in violent crime. This serves as an important validation of the empirical strategy and highlights the importance of looking beyond aggregate crime measures in this climate-crime literature, since they may obscure heterogeneous patterns across crime categories.

The Asian renaissance of the second half of the twentieth century has been primarily associated with substantial gains in agricultural output and productivity. However, that was not always the case as there were fears in the 1950s and the early 1960s that the tropical Asian rice-based economy would be experiencing massive famine and starvation because the region had already reached its cultivation frontier ( Otsuka and Larson 2013 ). Prior to the Green Revolution, food shortages and famines were considered a typical Asian phenomenon ( Gráda 2009 ). Nutritional intakes for the mass of the population did not improve, and mortality rates were considerably high in that part of the world ( Booth 2012 ). The results contained herein add supporting evidence to this idea, since it is found that falling agricultural incomes, and by extension rural poverty, primarily affected the categories of crime that alleviated economic distress. Beyond improving our understanding on local conditions of early twentieth century South and South-East Asian states, the implication of this study may be important from a public policy perspective in contemporary developing countries. Taken together, the results of this research support the idea of improved high-yield weather-resistant grains and investments in irrigation and drainage technology. The promise of a stable annual harvest would potentially eliminate much of the adverse crime-induced poverty traps, as well as the subsequent unfolding vicious cycle of crime and further disruptions to human welfare ( Bourguignon 2000 ). Unfortunately, climate change continues and it is going to bring about more erratic weather events, hitting the poorest smallholder farmers the most. A key policy priority should therefore be to aim at a long-term protection of the most vulnerable and precarious farmers of the global south.

Supplementary material are available at European Review of Economic History online .

Both types of water management systems are essential in tropical agriculture. After a flood, especially in areas that are flat and low-lying, water stagnates upon the soil rotting and eventually destroying the roots of the plants. Many scholars have pointed out that in this context “a system of controlled drainage is more important than irrigation, while a combination of both is the ideal” ( Lim 1976 , p. 43). Artificial drainage is, thus, necessary to circumvent surface run-offs, waterlogging, and other phytopathological diseases that impede root growth.

Twenty-seven states and districts were included in the analysis. The states include, the three Straits Settlements of Singapore, Penang, and Malacca; the four Malay states of Selangor, Perak, Negeri Sembilan, and Pahang which form the Federated Malay States (F.M.S.); Johore, Kedah, Perlis, Kelantan, and Trengganu which collectively are known as the Unfederated Malay States (U.M.S.); North Borneo with its five administrative districts; the protectorate of Brunei; and Ceylon with its nine administrative districts. Sarawak had to be dropped due to lack of consistent data.

India, for example, was not included since rice was largely competing with wheat as the main staple crop and almost half of its land lies outside the tropics.

For instance, in British Malaya around two-thirds of total rice availability was supplied by imports, mainly from Thailand and Burma.

See Booth (2012) for a detailed discussion on fiscal spending variation across colonial systems in South East Asia.

We are aware of potential inadequacies and biases in the colonial sources (e.g., the possibility that the district officials would consistently under-report issues to look great to their superiors). In Section 2.2 , we were able to match several unfavourable weather conditions with widespread distress and related property crime, while we were able to identify a null effect on violent crimes. If this bias of under-reporting exists in the case of British Asia, and negative income shocks cause households to under-report crime, our estimates would underestimate the true causal impact of poverty on crime. Moreover, the functional use of the archival sources enabled us to obtain a more thorough understanding of the important mechanisms driving the relationship of interest, and to add a new layer of robustness by backing up the regression results.

For comparison purposes, Table A8 in the Appendix presents the results for violent crime following an identical structure.

In addition, there can be classical measurement error, which would lead to attenuation bias.

In results not reported, we find that vagrancy rates exhibit a similar outcome for both high and low percentage of per capita padi-rice cultivation. The effect was statistically significant in both subsamples, but it was not statistically distinguishable from each other (p-value = 0.344).

Anderson , C.A. ( 1989 ). Temperature and aggression: ubiquitous effects of heat on occurrence of human violence . Psychological Bulletin 106 , pp. 74 .

Google Scholar

Anderson , C.A. ( 2001 ). Heat and violence . Current Directions in Psychological Science 10 ( 1 ), pp. 33 – 38 .

Anderson , C.A. , Anderson , K.B. , Dorr , N. , DeNeve , K.M. and Flanagan , M. ( 2000 ). Temperature and aggression . Advances in Experimental Social Psychology 32 , pp. 63 – 129 .

Anderson , R.W. , Johnson , N.D. and Koyama , M. ( 2015 ). Jewish Persecutions and Weather Shocks: 1100‐1800 . The Economic Journal . Doi:10.1111/ecoj.12331 .

Bai , Y. and Kung , J.K.S. ( 2011 ). Climate shocks and sino-nomadic conflict . Review of Economics and Statistics 93 , pp. 970 – 81 .

Becker , G. ( 1968 ). Crime and punishment: An economic approach . The Journal of Political Economy 76 , pp. 169 – 217 .

Bignon , V. , Caroli , E. and Galbiati , R. ( 2016 ). Stealing to survive? Crime and income shocks in nineteenth century France . The Economic Journal .

Blakeslee , D. and Fishman , R. ( 2014 ). Weather shocks, crime, and agriculture: evidence from India . Social Science Research Network .

Blattman , C. and Miguel , E. ( 2010 ). Civil war . Journal of Economic Literature 48 , pp. 3 – 57 .

Blau , J.R. and Blau , P.M. ( 1982 ). The cost of inequality: Metropolitan structure and violent crime . American Sociological Review , pp. 114 – 129 .

Booth , A. ( 1999 ). Initial conditions and miraculous growth: why is South East Asia different from Taiwan and South Korea . World Development 27 ( 2 ), pp. 301 – 321 .

Booth , A. ( 2012 ). Measuring living standards in different colonial systems: some evidence from South East Asia, 1900–1942 . Modern Asian Studies 46 ( 05 ), pp. 1145 – 1181 .

Bourguignon , F. ( 2000 , December). Crime, violence and inequitable development. In Annual World Bank Conference on Development Economics 1999 (pp. 199–220).

Bray , F. ( 1994 ). The Rice Economies: Technology and Development in Asian Societies . Berkeley, USA : University of California Press .

Google Preview

Brückner , M. and Ciccone , A. ( 2011 ). Rain and the democratic window of opportunity . Econometrica 79 ( 3 ), pp. 923 – 947 .

Bruton , H.J. ( 1992 ). The Political Economy of Poverty, Equity, and Growth: Sri Lanka and Malaysia . Oxford University Press .

Buhaug , H. , Nordkvelle , J. , Bernauer , T. , Böhmelt , T. , Brzoska , M. , Busby , J.W. , .. and Goldstone , J.A. ( 2014 ). One effect to rule them all? A comment on climate and conflict . Climatic Change 127 ( 3–4 ), pp. 391 – 397 .

Burgess , R. and Donaldson , D. ( 2010 ). Can openness mitigate the effects of weather shocks? Evidence from India's famine era . The American Economic Review 100 ( 2 ), pp. 449 – 453 .

Burke , M. , Dykema , J. , Lobell , D.B. , Miguel , E. and Satyanath , S. ( 2015 ). Incorporating climate uncertainty into estimates of climate change impacts . Review of Economics and Statistics 97 , pp. 461 – 71 .

Butcher , G.J. ( 1979 ). The British in Malaya, 1880– 1941 . New York : Oxford University Press .

Christian , C. , & Fenske , J. ( 2015 ). Economic shocks and unrest in French West Africa (No. 2015-01). Centre for the Study of African Economies, University of Oxford.

Dell , M. ( 2012 ). Path dependence in development: Evidence from the Mexican Revolution. Harvard University, mimeograph. ISO 690

Dell , M. , Jones , B.F. and Olken , B.A. ( 2014 ). What do we learn from the weather? The new climate–economy literature . Journal of Economic Literature 52 ( 3 ), pp. 740 – 798 .

Drabble , J.H. ( 1973 ). Rubber in Malaya 1876–1922: The Genesis of the Industry . Kuala Lumpur; New York : Oxford University Press .

Elson , R.E. ( 1997 ). The End of the Peasantry in Southeast Asia: A Social and Economic History of Peasant Livelihood, 1800-1990s . London : Macmillan Press LTD .

Fajnzylber , P. , Lederman , D. and Loayza , N. ( 2002 ). What causes violent crime . European Economic Review 46 ( 7 ), pp. 1323 – 1357 .

Farmer , B.H. ( 1977 ). Pioneer Peasant Colonization in Ceylon . London : Oxford University Press .

Fenske , J. and Kala , N. ( 2015 ). Climate and the slave trade . Journal of Development Economics 112 , pp. 19 – 32 .

Fjelde , H. and von Uexkull , N. ( 2012 ). Climate triggers: rainfall anomalies, vulnerability and communal conflict in sub-Saharan Africa . Political Geography 31 , pp. 444 – 53 .

González , F. and Miguel , E. ( 2015 ). War and local collective action in Sierra Leone: a comment on the use of coefficient stability approaches . Journal of Public Economics 128 , pp. 30 – 3 .

Gráda , C.Ó. ( 2009 ). Famine: A Short History . Princeton University Press .

Grossman , H.I. ( 1991 ). A general equilibrium model of insurrections. The American Economic Review 81 , pp. 912 – 21 .

Herbst , J. ( 2014 ). States and Power in Africa: Comparative Lessons in Authority and Control . Princeton University Press .

Hill , R.D. ( 2012 ). Rice in Malaya: A Study in Historical Geography . Singapore : NUS Press .

Hoechle , D. ( 2007 ). Robust standard errors for panel regressions with cross-sectional dependence . Stata Journal 7 , pp. 281 .

Hsiang , S.M. , Burke , M. and Miguel , E. ( 2013 ). Quantifying the influence of climate on human conflict . Science 341 , 1235367 .

Iyer , L. , and Topalova , P.B. ( 2014 ). Poverty and crime: evidence from rainfall and trade shocks in India. Harvard Business School BGIE Unit Working Paper (14–067).

Jia , R. ( 2014 ). Weather shocks, sweet potatoes and peasant revolts in historical China . The Economic Journal 124 , pp. 92 – 118 .

Kelly , M. ( 2000 ). Inequality and crime . Review of Economics and Statistics 82 ( 4 ), pp. 530 – 539 .

Klomp , J. and Bulte , E. ( 2013 ). Climate change, weather shocks, and violent conflict: a critical look at the evidence . Agricultural Economics 44 , pp. 63 – 78 .

Lange , M.K. ( 2004 ). British colonial legacies and political development . World Development 32 ( 6 ), pp. 905 – 922 .

Lim , T.G. ( 1976 ). Origins of a Colonial Economy: Land and Agriculture in Perak , 1874 – 1879 . Penerbit University Sains Malaysia : distributed by Federal Publications .

Matsuura , K. and Willmott , C.J. ( 2009 ). Terrestrial precipitation: 1900–2010 gridded monthly time series.

Mehlum , H. , Miguel , E. and Torvik , R. ( 2006 ). Poverty and crime in 19th century Germany . Journal of Urban Economics 59 , pp. 370 – 88 .

Miguel , E. ( 2005 ). Poverty and witch killing . The Review of Economic Studies 72 , pp. 1153 – 72 .

Miguel , E. , Satyanath , S. and Sergenti , E. ( 2004 ). Economic shocks and civil conflict: an instrumental variables approach . Journal of Political Economy 112 , pp. 725 – 53 .

Mills , L.A. ( 1964 ). Ceylon Under British Rule . New York, Barnes & Noble : Routledge , pp. 1795 – 932 .

Oster , E. ( 2014 ). Unobservable selection and coefficient stability: Theory and evidence. University of Chicago Booth School of Business Working Paper.

Otsuka , K. and Larson , D.F. ( 2013 ). Towards a green revolution in Sub-Saharan Africa. In An African Green Revolution . Netherlands : Springer , pp. 281 – 300 .

Papaioannou , K.J. ( 2016 ). Climate shocks and conflict: evidence from colonial Nigeria . Political Geography 50 , pp. 33 – 47 .

Papaioannou , K.J. and DeHaas , M. ( 2017 ). Weather Shocks and Agricultural Commercialization in Colonial Tropical Africa: Did Export Crops Alleviate Social Distress? World Development . Doi: 10.1016/j.worlddev.2017.01.019 .

Parmer , N.J. ( 1960 ). Colonial Labor Policy and Administration: A history of labor in the rubber plantation industry in Malaya , 1910 – 1941 . Locust Valley, N.Y. : Austin .

Patterson , E.B. ( 1991 ). Poverty, income inequality, and community crime rates . Criminology 29 ( 4 ), pp. 755 – 776 .

Ranson , M. ( 2014 ). Crime, weather, and climate change . Journal of Environmental Economics and Management 67 , pp. 274 – 302 .

Roeder , O.K. , Eisen , L.B. , Bowling , J. , Stiglitz , J.E. and Chettiar , I.M. ( 2015 ). What caused the crime decline? Columbia Business School Research Paper (15–28).

Sarsons , H. ( 2015 ). Rainfall and conflict: a cautionary tale . Journal of Development Economics 115 , pp. 62 – 72 .

Seter , H. ( 2016 ). Connecting climate variability and conflict: implications for empirical testing . Political Geography 53 , pp. 1 – 9 .

Tregonning , K.G. ( 1965 ). A History of Modern Sabah (North Borneo, 1881–1963). University of Singapore.

Wade , R. ( 1998 ). The Asian debt-and-development crisis of 1997-?: Causes and consequences . World Development 26 ( 8 ), pp. 1535 – 1553 .

All documents were retrieved from The National Archives, London. Government Printing Office: various issues, reported in alphabetical order.

Brunei Annual Report: Department code: CO 842, Series number: 1–2.

Ceylon Administration Report: Department code: CO 57, Series number: 176–258.

Ceylon Blue Book: Department code: CO 59, Series number: 120–152.

Federated Malay States Administration Report: Department code: CO 576, Series number: 5–65.

Federated Malay States Blue Book: Department code: CO 575, Series number: 1–37.

Federated Malay States Provincial Report: Department code: CO 435/437/438/439, Series number: 1–4.

Johore Annual Report: Department code: CO 715, Series number: 1–5.

Kedah & Perlis Annual Report: Department code: CO 716, Series number: 1–4.

Kelantan Annual Report: Department code: CO 827, Series number: 1-2.

North Borneo (Sabbah) Annual Report: Department code: CO 648, Series number: 1–22.

North Borneo (Sabbah) Blue Book: Department code: CO 966, Series number: 1.

Straits Settlements Annual Report: Department code: CO 275, Series number: 84–151.

Straits Settlements Blue Book: Department code: CO 277, Series number: 53–91.

Trengganu Annual Report: Department code: CO 840, Series number: 1-2.

Supplementary data

Email alerts, citing articles via.

• The EHES Blog
• Recommend to your Library

Affiliations

• Online ISSN 1474-0044
• Print ISSN 1361-4916
• Copyright © 2024 European Historical Economics Society
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.