![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
Despite decades of research exploring the relationship between the economy and crime, there is a lack of clarity in this literature. Questions remain, particularly with respect to how the economy is measured and whether the relationship between the economy and crime is the same across contexts. The literature to date has overlooked what is called the “shadow” economy, which represents unreported economic activity. We examine the relationship between traditional measures of the legitimate economy (e.g. unemployment) and crime as well as whether the shadow economy moderates this relationship using a panel of U.S. states from 1997 to 2008. Our results show that the shadow economy reduces the strength of the relationship between the legitimate economy and crime, and the effect of the economy on crime is conditional on the size of the shadow economy.
Keywords:
Notes
1 Note, in the remainder of this paper, we refer to the shadow economy as unrecorded legitimate activity, as this is what it is intended to capture. Wiseman (Citation2013, p. 312) states that the MIMIC method used in this paper “focuses more narrowly on otherwise legal market activity” than other measures of the shadow economy. These activities can also be thought of as part of the “informal” or “unofficial” economy. We primarily use the term shadow economy throughout to reflect the language used in Wiseman (Citation2013).
2 Examples of the shadow economy's existence and prevalence in the U.S. has recently gained significant popularity in the media (see, e.g. Newman, Citation2013; Plumer, Citation2013; Surowiecki, Citation2013).
3 It is important to point out that neither of these studies were published in a peer reviewed, academic venue.
4 Some argue that the presence of the shadow economy causes official statistics to give false signals (see Feige, Citation1989).
5 Rosenfeld and Messner (Citation2013) find that the relationship between crime (particularly, violent crime) and the economy is moderated by extensive employment insurance, health insurance, and pensions.
6 Goel and Saunoris (Citation2016) and Wiseman (Citation2015) also use this measure of state-level shadow activity.
7 Instrumental crime refers to crimes undertaken for tangible economic (financial, property) gain, including robbery, theft, and burglary.
8 Note that we do not include a formal hypothesis regarding the relationship between the legitimate economy and crime absent the shadow economy because of previous inconsistencies. Our argument is that prior econometric models have been misspecified by omitting the shadow economy as an explanatory variable.
9 The various dependent variables allow us to examine the relationships of interest, while recognizing that the direction and significance of the relationship may depend on the type of crime. These data are collected from the FBI’s Uniform Crime Report, available at http://www.ucrdatatool.gov/.
10 For a useful survey related to the challenges of estimating the shadow economy, see Frey and Weck-Hanneman (Citation1984).
11 Schneider and Enste (Citation2013) discuss twelve distinct techniques for estimating the shadow economy in the literature.
12 The Hausman test was used to determine whether fixed effects or random effects estimation was more appropriate. For all of the dependent variables, except larceny, we reject the null hypothesis, implying that random effects are not consistent. Therefore, to maintain consistency and because fixed effects is consistent under both hypotheses, we proceed with fixed effects estimation.
13 We considered an alternate model using generalized least squares (GLS) and population weights assuming first-order autocorrelation and a heteroskedastic error structure. Overall, our main results remain robust to this alternate specification.
14 We choose to log the dependent variable following Levitt (Citation2001).
15 In all specifications, robust standard errors, based on the Huber-White sandwich estimator, are used to ensure the standard errors are robust to heteroscedasticity and other forms of misspecification.
16 All variables were tested to determine whether they are stationary. The Levin–Lin–Chu (Levin, Lin, & James Chu, Citation2002) test was used to test for unit roots. We reject the null hypothesis that the series contains a unit root for all of the variables except population density. If we include the first difference of population density instead of the level, the results presented below are robust.
17 Within the shadow economy literature, the effect of unemployment on the shadow economy is ambiguous (see Buehn & Schneider, Citation2012).
18 Introducing an interaction term into the model has the potential to create problems with multicollinearity; however, the large sample size mitigates these concerns. Furthermore, the omitted variable bias generated by the omission of the interaction term outweighs the problems associated with inflated estimator variances as evidenced by the retained statistical significance of our results and intuition of the non-linear pattern predicted.
19 It is important to note that if the correct specification identifies a statistically significant relationship between unemployment and crime conditional on the level of the shadow economy, we cannot interpret the results in Tables and in any meaningful way, as they suffer from potential misspecification and omitted variable bias.
20 The existing literature finds that property crime is an important predictor of violent crime and homicide (Rosenfeld, Citation2009). We ran an alternate specification including property crime as a regressor in the models for violent crime and homicide. While the coefficient of property crime is statistically significant and positive, our main findings are not significantly affected. The results are available upon request.
21 In addition, we consider inclusion of two other control variables that have been shown to be important determinants of crime in the literature: divorce rates and immigration (legal and illegal immigration were examined separately). Augmenting regression specification (3) with these variables, both separately and together, does not significantly influence our main results. We also conduct a robustness check that includes a lagged unemployment term and an interaction term between the shadow economy measure and lagged unemployment. Again, our main results remain robust. The results are available upon request. Finally, as an additional robustness check, and to account for possible cross-sectional dependence we re-estimated our models using Pesaran’s (Citation2006) Common Correlated Effects estimator. This estimator models cross-sectional dependence using unobserved common factors with heterogeneous factor loadings. Although this estimator relies on medium to large N and T, the shadow economy continues to moderate the effect of unemployment on crime rates as illustrated by the modeling of the effect of the economy at different levels of the shadow economy. Results are available upon request.
22 These results are consistent with the findings of Phillips and Land (Citation2012). Phillips and Land (Citation2012) examine the effect of unemployment on crime at the county, state, and national levels and find that for violent crimes, the relationship with the unemployment rate is weak and not statistically significant, but there is a significant relationship between unemployment burglary, larceny, and motor vehicle theft. Our results support this conclusion.
23 Many previous researchers have used the MIMIC model as a way of estimating the size of the shadow economy (see, for example, Dell’Anno & Solomon, Citation2008; Frey & Weck-Hanneman, Citation1984; Schneider, Buehn, & Montenegro, Citation2010).
24 For an illustrative description of the MIMIC model’s general structure, see Figure 2 in Wiseman (Citation2013). For a detailed description of the MIMIC method, consult Schneider and Buehn (Citation2013).
25 Wiseman (Citation2013) considers other potential causal variables including the unemployment rate and sole-proprietorship as a percent of total employment, but these variables were not included in the main specification.
Additional information
Notes on contributors
Michael Rocque
Michael Rocque is an assistant professor of Sociology at Bates College. His research interests include desistance from crime and the demography of crime.
James W. Saunoris
James W. Saunoris is an associate professor of Economics at Eastern Michigan University. His research interests include the shadow economy, corruption, and entrepreneurship.
Emily C. Marshall
Emily C. Marshall is an assistant professor of Economics at Dickinson College. Her research interests are primarily in the field of macroeconomics and monetary policy, specifically, examining the impact of several different housing market features on macroeconomic volatility.