The impact of natural resources on crime rates: Global evidence since 1890 shows how silver mining increased homicide rates

Jessica Baier, Jörg Baten 19 November 2017



Recent upswings in violent crime across both North and South America has sparked a debate on the driving factors and determinants of crime (Abt 2017). Scholars from many scientific fields have already spent decades trying to explain why some regions are violent and insecure, while others have remained relatively peaceful. Inequality, poverty, weak law enforcement, low educational levels, and cultural and social fractionalisation have all been blamed for contributing to high crime rates (Fajnzylber et al. 1998, Buonanno 2003).

Natural resources such as precious metals, minerals, diamonds, or oil have been related to violent civil war and conflict, but their link to rates of violent crime during peacetime had have not been explored. Couttenier et al. (2014) related the extraction of valuable mining resources to an increase in homicide rates by examining the US gold rush of the mid-19th century. If gold discoveries happened before formal institutions were in place, they found that interpersonal violence tended to increase particularly strongly.

One hypothesis is that weak institutional setups mean there is weak enforcement of property rights. Violence becomes a substitute for the public enforcement of property rights. When there are valuable resources, the need for property rights enforcement increases, as does violence. Violence is used for direct expropriation and defence of property, as well as for pre-emption. Another potential channel is that in weak institutional setups, organisations prone to violence (gangs, guerrillas, paramilitary groups) benefit from the additional revenue generated by the resources, instead of the government. This allows them to become more violent and powerful, purchasing more weapons, and exerting more influence.

Capturing peacetime homicide rates

In recent research, we adopt a global and long-term perspective (Baier and Baten 2017). We collect and code data on homicide rates, supplemented with publicly available sources (WHO mortality database, ClioInfra, Historical Violence Database, Interpol) to construct a database that covers more than 50 countries from 1890 until 2000. The evidence includes data on Africa, Latin America and Asia.

We use the homicide rate per 100,000 inhabitants as an indicator of interpersonal violence. Homicide rates have the advantage that there are fewer definition issues than statistics such as non-lethal violence, which has many different definitions.

Early research by social historians and legal experts, and UN organisations more recently, allows us to calculate homicide rates worldwide, even for early periods in the data. Estimating homicide rates across 110 years has its problems (discussed in our paper), but it is an important insight into regional trends and differences in interpersonal violence.

We use the share of GDP that was produced by silver mining (silver dependence) as an indicator of the presence of valuable resources. Silver can serve as a resource indicator over a long period as well as a global sample:

  • Silver is an important mining product for many countries, and can be found in all world regions, whereas gold, diamonds, or precious metals such as antimony are more localised, and so only important for small groups of countries.
  • The law of one price is also not strongly violated for silver, because it is easy to transport.
  • Technological change matters for silver, but it less than for other resources like oil or guano.

A glance at Figure 1 suggests that silver dependence and homicide rates might be related. It shows the trends for Mexico and Peru from 1920 until 2010.

Figure 1 Homicide rates and silver dependence in Mexico (Panel A) and Peru (Panel B)

Panel A: Mexico

Panel B: Peru

Source: Baier and Baten (2017).

  • For Mexico, there was a small increase in silver dependence between 1920 and 1930 as the economy recovered from political conflicts during the Mexican revolution. After 1940, there was a consistent decline in silver dependence. The homicide rate tracked this change.
  • For Peru, we can see both decreases and increases in silver dependence and murder. The initial decline in silver dependence around the 1930s was followed by an increase in the 1960s, and a climax around the 1990s, in both silver dependence and homicide.

To test the hypothesis that silver dependence increased homicide rates empirically, we use decadal averages to rule out short-term fluctuations.

First, we estimate fixed effects and correlated random effects. We include control variables, including GDP per capita. Low GDP levels might proxy a low opportunity cost (it makes it easier to recruit young men or criminal gangs) but also an undiversified economic structure in which high silver dependence could also measure economic monoculture. By controlling for GDP per capita, we take this effect into account.

Inequality is another important control variable. Criminal gangs often appear when assets can be transferred from rich to poor in high-inequality settings. We also control for education, drug production and trade, violent conflicts, and other factors. The results suggest that, for the global dataset, higher silver dependence is related to an increase in homicide rates.

The scale of the resource curse

We use instrumental variables to make sure that the results are not affected by endogeneity due to omitted variables or reverse causality. By altering the incentive structure in an economy, homicide rates might increase or reduce silver production. We use the world market price of silver and the number of sites in which silver occurs, per country, relative to country size as instrumental variables. Both variables are sources of exogenous variation that influence silver production in a country, but they are not determined by characteristics of violence in any single country. The effect is even more pronounced when integrating these exogenous sources of variation, and so the previous results are not inflated by reverse causality or an omitted variable, but rather the effect might even be underestimated.

The 'resource curse', then, has increased interpersonal violence in peacetime periods in more than 50 countries over the last 110 years. Further research can determine if these results hold for other resource types, and if policy interventions can offset the negative effect of mining resources on interpersonal violence.


Abt, T (2017), "How Not to Respond to the Rising Murder Rate", The New York Times, 26 September.

Baier, J and J Baten (2017), "Silver, Murder, and Institutions: Did the 'Curse of Resources' impact on Homicide Rates? Global evidence since 1890", CEPR Discussion Paper No. DP12397.

Buonanno, P (2003), "The Socioeconomic Determinants of Crime. A Review of the Literature", University of Milano-Bicocca, Department of Economics working paper No. 63.

Couttenier, M, P Grosjean and M Sangnier (2014), "The Wild West is Wild: The Homicide Resource Curse", The University of New South Wales School of Economics discussion paper No. 2014–12.

Fajnzylber, P, D Lederman and N Loayza (1998), Determinants of crime rates in Latin America and the world: an empirical assessment, The World Bank.



Topics:  Development Economic history

Tags:  resource curse, violence, homicide, mining, silver, Mexico, Peru

PhD candidate, University of Tuebingen

professor of Economic History, University of Tuebingen