Abstract
Objective
This paper uses a “local average treatment effect” (LATE) framework in an attempt to disentangle the separate effects of criminal and noncriminal gun prevalence on violence rates. We first show that a number of previous studies have failed to properly address the problems of endogeneity, proxy validity, and heterogeneity in criminality. We demonstrate that the time series proxy problem is severe; previous panel data studies have used proxies that are essentially uncorrelated in time series with direct measures of gun relevance.
Methods
We adopt instead a cross-section approach: we use US county-level data for 1990, and we proxy gun prevalence levels by the percent of suicides committed with guns, which recent research indicates is the best measure of gun levels for crosssectional research. We instrument gun levels with three plausibly exogenous instruments: subscriptions to outdoor sports magazines, voting preferences in the 1988 Presidential election, and numbers of military veterans. In our LATE framework, the estimated impact of gun prevalence is a weighted average of a possibly negative impact of noncriminal gun prevalence on homicide and a presumed positive impact of criminal gun prevalence.
Results
We find evidence of a significant negative impact, and interpret it as primarily “local to noncriminals”, i.e., primarily determined by a negative deterrent effect of noncriminal gun prevalence. We also demonstrate that an ATE for gun prevalence that is positive, negative, or approximately zero are all entirely plausible and consistent with our estimates of a significant negative impact of noncriminal gun prevalence.
Conclusions
The policy implications of our findings are perhaps best understood in the context of two hypothetical gun ban scenarios, the first more optimistic, the second more pessimistic and realistic. First, gun prohibition might reduce gun ownership equiproportionately among criminals and noncriminals, and the traditional ATE interpretation therefore applies. Our results above suggest that plausible estimates of the causal impact of an average reduction in gun prevalence include positive, nil, and negative effects on gun homicide rates, and hence no strong evidence in favor of or against such a measure. But it is highly unlikely that criminals would comply with gun prohibition to the same extent as noncriminals; indeed, it is virtually a tautology that criminals would violate a gun ban at a higher rate than noncriminals. Thus, under the more likely scenario that gun bans reduced gun levels more among noncriminals than criminals, the LATE interpretation of our results moves the range of possible impacts towards an increase in gun homicide rates because the decline in gun levels would primarily occur among those whose gun possession has predominantly negative effects on homicide.
Similar content being viewed by others
Notes
See, e.g., Sloan et al. (1990), Killias (1993), and Zimring and Hawkins (1997). Detailed studies using cross-national data are, however, generally unsupportive of this conclusion, and suggest instead that there is no significant association between national gun ownership rates and rates of homicide, suicide, robbery, or assault (Kleck 1997, p. 254; Killias et al. 2001, pp. 436, 440).
See our 2005 CEPR discussion paper for a discussion and list of 30 such studies, and Kleck (1997), Chapter 7 for a more extensive review of the pre-1997 research.
Greene (2008), pp. 29–30 has a short and clear discussion of how to interpret the partial R2 for an OLS regression. Note that the “within R2” for the fixed effects model can be interpreted as the partial R2 after the panel dummies have been partialed out.
The original dataset is kindly provided on Moody’s website, http://cemood.people.wm.edu/research.html.
“The significantly positive estimate of 0.354 suggests that this magazine’s sales are a valid measure both of the level and of the change in gun ownership within an area.” (Duggan 2001, p. 1093) We note that his claim that this regression is evidence that GAR is correlated with the level of g is also mistaken; a FE specification cannot justify cross-sectional validity. But the mistake of confusing statistical significance with proxy validity is the key error.
Duggan reports a coefficient of 0.354 with a standard error of 0.114, giving a t-statistic of 3.105. The estimation has 488 observations with 45 state dummies and 13 estimated parameters (log(HHG) and 12 year dummies), leaving 430 residual degrees of freedom. Using the formula provided in the previous section, the implied partial R2 is t2/(t2 + df) = 3.1052/(3.1052 + 430) = 2 %.
The t-statistics for the significance of PSG in these two regressions are 1.108/0.417 = 2.657 and 0.905/0.355 = 2.549, respectively. The corresponding partial R2s are therefore 2.6572/(2.6572 + 103) = 6 % and 2.5492/(2.5492 + 117) = 5 %.
We use the answer to the BRFSS survey question, “Are any firearms now kept in or around your home?” We exclude respondents who answered “don’t know/not sure” or who refused to answer. Non-respondents account for less than 5 % of the total sample in any year. We used the survey weights provided with the BRFSS; the unweighted measures generate very similar results.
We are distinguishing here between (a) “criminals” as a type of person, and (b) “criminality” or “criminal behavior”. The terms ‘criminal’ and ‘noncriminal’ are, of course simplifications, and can be regarded as shorthand for “persons who commit serious crimes” and “persons who do not commit serious crimes.
For example, in 2000 there were 16,765 total homicides counted in the vital statistics data for the US, but only 137 civilian justifiable homicides with firearms (vital statistics homicide counts include justifiable homicides by civilians but exclude those by police), or 8/10 of one percent of total homicides (US Federal Bureau of Investigation 2000, p. 24).
The use of guns by criminals for self-defense against other criminals is likely to have a negative effect, but is probably insufficient to offset the positive criminality effects.
This interpretation occasionally appears in textbooks and expositions of IV/GMM. Deaton (1997, p. 112) provides a good example: “the OID [overidentifying restrictions] test tells us whether we would get (significantly) different answers if we used different instruments or different combinations of instruments in the regression. … If we have only k instruments and k regressors, the model is exactly identified, … there is only one way of using the instruments, and no alternative estimates to compare.”
Occasionally one finds in the literature the claim is that a test of overidentifying restrictions has power only if there is a subset of instruments that are all valid and identify the model. The claim is incorrect; the correct statement is that the test will have power if there is such a subset, and might, or might not, lack power, if there are not enough valid instruments to identify the model.
White (1994) is very clear on this point, for example. “Nor is there any necessity for either estimator to retain consistency in the presence of misspecification. Power is achieved because the estimators chosen have differing probability limits under misspecified alternatives. These alternatives necessarily go beyond those that allow one of the estimators to retain consistency for a certain parameter value.” White (1994), p. 274.
The 1988 election results were chosen in preference to the 1992 results because the date precedes the census year from which most our data are taken (and hence is more plausibly exogenous), and because the choice between the two main candidates in 1988 maps more closely to attitudes towards gun ownership: in the 1992 election, unlike the 1988 election, the politically less conservative candidate (negatively correlated with gun ownership) was also a southerner (positively correlated with gun ownership). The 1992 results are also less easily interpreted because of the significant share of the vote that went to the third-party candidate, Ross Perot.
An anonymous reviewer also questioned our use of the voting Republican instrument because of reverse causality. The extant evidence, however, only supports an effect of publicity about crime (usually concerning especially notorious crimes) on political preference, not changes in actual crime rates. Since studies have repeatedly found publicity (news coverage) about crime to be unrelated to actual crime rates, the evidence about the effects of publicized crime has no bearing on the effects of actual crime rate on voting preferences. We are not aware of any evidence indicating that actual crime rates affect voting preferences.
See, e.g., the case control study by Kleck and Hogan (1999, p. 285), who found that veterans are 2.8 times more likely than nonveterans to commit murder.
Contra Cook and Ludwig (2003b, p. 12): “the usual approach [to addressing heterogeneity in cross-sectional gun/crime studies] … has been to statistically control for the handful of local characteristics that are readily available in standard data sources, such as population density, poverty, and the age and racial composition of the population. But these variables never [sic] explain very much of the cross-sectional variation in crime rates, suggesting that the list of control variables is inadequate to the task.” Our results suggest this view is too pessimistic about the feasibility of cross-sectional studies.
Stock and Yogo do not tabulate critical values for the non-homoskedastic case. We use the IV critical values on the grounds that the IV estimator is a special case of two-step GMM for homoskedastic and independent errors. Using the simple alternative of the Staiger and Stock (1997) rule of thumb that the F statistic should be at least 10 to avoid weak instrument problems leads to similar conclusions as those we report below.
The results are similar when using the traditional but inefficient IV estimator.
The latter three robustness checks derive from comments by two anonymous referees.
The IV estimator is used when there is only one excluded instrument because the 2-step GMM estimator reduces to IV in the just-identified case.
An exogeneity test of PSG using G&A as the sole instrument yields a χ2(1) statistic of 0.28 with a p value of 0.600. As noted earlier, this is equivalent to a test of the difference between an estimate of b1 treating PSG as exogenous and an estimate using G&A as an instrument for PSG.
Obtained by differentiating Eq. (5) with respect to g C, noting that h is a log crime rate, and then multiplying both sides by \( \bar{g}^{C} \) to obtain the elasticity at the mean.
The differences arise because of the non-overlap of the small number of missing values for the two dependent variables.
These are the county means of the shares of gun and nongun homicide in total homicide for the total sample of 1,456 counties.
That is, we double the size of the dataset so that all counties appear twice, once where the dependent variable is the log gun homicide rate and again where it is the log nongun homicide rate. We then interact all regressors and instruments, including the fixed effects, with a gun/nongun homicide dummy, so that all of the variables in the gun homicide equation take the value of zero when the dependent variable is nongun homicide, and visa-versa for the nongun homicide equation variables.
This works because clustering on states allows for arbitrary within-state correlations. This includes the possible correlation between the two observations on an individual county (one from the gun equation, one from the nongun equation).
http://ideas.repec.org/SoftwareSeries.html. ivreg2 is a general-purpose IV/GMM estimation routine for linear models; xtivreg2 supports fixed-effects panel data models.
If X and Y are two random variables, then \( {\text{E}}\left( \frac{Y}{X} \right) = \frac{{{\text{E}}(Y)}}{{{\text{E}}(X)}} - \frac{{{\text{cov}}(X,\frac{Y}{X})}}{{{\text{E}}(X)}} \) (Frishman 1971, p. 333, Eq. 4.2), provided that E(X) ≠ 0, i.e., we are assuming that the mean of total gun prevalence is nonzero. The derivation of Eq. (27) also makes use of the fact that \( \text{cov} \left( {g_{i} ,\frac{{g_{\text{i}}^{\text{NC}} }}{{g_{i} }}} \right) = \text{cov} \left( {g_{i} ,1 - \frac{{g_{\text{i}}^{\text{C}} }}{{g_{i} }}} \right) = - \text{cov} \left( {g_{i} ,\frac{{g_{\text{i}}^{\text{C}} }}{{g_{i} }}} \right) \). Note that the weights on \( {{\upbeta}}_{ 1}^{\text{C}} \) and \( {{\upbeta}}_{ 1}^{\text{NC}} \) in (27) sum to 1.
This is the ICPSR dataset 2266, “Directory of Law Enforcement Agencies: 1992”. We use total sworn police, aggregating agencies by county. We exclude agencies with state-wide jurisdiction. We do not calculate the measure for the 5 counties of New York City, because sworn police for NYC are not available in the dataset disaggregated by county. The per capita measure uses county population according to the 1990 census.
The measure is constructed using the ICPSR datasets 9573, 9785 and 6036, “Uniform Crime Reporting Data [United States}: County-level Detailed Arrest and Offense Data”. The numerator uses data on arrests (all ages) for violent crimes; the denominator uses the sum of reported murders, rapes, robberies and aggravated assaults. We aggregate over 1989–91 because of small numbers for some counties.
References
Angrist J, Imbens G (1995) Two-stage least squares estimates of average causal effects in models with variable treatment intensity. J Am Stat Assoc 90(430):431–442
Annest JL, Mercy JA, Gibson DR, Ryan GW (1995) National estimates of nonfatal firearm-related injuries. J Am Med Assoc 273:1749–1754
Audit Bureau of Circulations (1993) Supplementary data report, covering county paid circulation for gun and related sports magazines. Audit Bureau of Circulations, Schaumburg, IL
Ayres I, Donohue JJ III (2003) Shooting down the ‘More Guns, Less Crime’ hypothesis. Stanf Law Rev 55:1193–1312
Azrael D, Cook PJ, Miller M (2004) State and local prevalence firearms ownership: measurement structure, and trends. J Quant Criminol 20:43–62
Baum CF, Schaffer ME, Stillman S (2003) Instrumental variables and GMM: estimation and testing. Stata J 3:1–31
Baum CF, Schaffer ME, Stillman S (2007) Enhanced routines for instrumental variables/generalized method of moments estimation and testing. Stata J 7:465–506
Baum C, Schaffer ME, Stillman S (2008) IVREG2: Stata module for extended instrumental variables/2SLS and GMM estimation. http://ideas.repec.org/c/boc/bocode/s425401.html
Bordua DJ, Lizotte AJ (1979) Patterns of legal firearms ownership: a cultural and situational analysis of Illinois counties. Law Policy Quart 1:147–175
Clarke RV, Mayhew P (1988) The British gas suicide story and its criminological implications. In: Tonry M, Morris N (eds) Crime and justice, vol 10. University of Chicago Press, Chicago, pp 79–116
Cook PJ, Ludwig J (1997) Guns in America. Police Foundation, Washington, DC
Cook PJ, Ludwig J (2003a) Guns and burglary. In: Ludwig J, Cook PJ (eds) Evaluating gun policy. Brookings Institution Press, Washington, DC
Cook PJ, Ludwig J (2003b) Pragmatic gun policy. In: Cook and Ludwig (2003a), op. cit
Cook PJ, Ludwig J (2004) The social costs of gun ownership. NBER working paper 10736. http://www.nber.org/papers/w10736
Cook PJ, Ludwig J (2006) The social costs of gun ownership. J Public Econ 90:379–391
Deaton A (1997) The analysis of household surveys: a microeconometric approach to development policy. World Bank/Johns Hopkins University Press, Washington, DC
Decker SH, Curry GD, Catalano S, Watkins A (2005) Strategic approaches to Community Safety Initiative (SACSI) in St. Louis. Final Report. Research report submitted to the US Department of Justice, Document Number 210361, Award Number 2000-IJ-CX-K008, June. http://www.ncjrs.gov/pdffiles1/nij/grants/210361.pdf
Duggan M (2001) More guns, more crime. J Polit Econ 109:1086–1114
Duggan M (2003) Guns and suicide. In: Cook and Ludwig (2003a), op. cit
Frishman F (1971) On the arithmetic means and variances of products and ratios of random variables. Army Research Office, Durham, North Carolina, publication AD-785 623. http://handle.dtic.mil/100.2/AD0785623
Greene, WH (2008) Econometric analysis, 6th edn. Prentice Hall, New Jersey
Hayashi F (2000) Econometrics. Princeton University Press, Princeton
Heckman J (1997) Instrumental variables: a study of implicit behavioral assumptions used in making program evaluations. J Human Resourc 32(3):441–462
ICPSR (1995) General election data for the United States, 1950–1990 [Computer file]. ICPSR ed. Inter-university Consortium for Political and Social Research [producer and distributor], Ann Arbor, MI
Imbens G, Angrist J (1994) Identification and estimation of local average treatment effects. Econometrica 62(2):467–475
Kates DB, Mauser G (2007) Would banning firearms reduce murder and suicide? A review of international and some domestic evidence. Harv J Law nd Public Policy 30(2):649–694
Killias M (1993) Gun ownership, suicide, and homicide: an international perspective. In: del Frate A, Zvekic U, van Dijk JJM (eds) Understanding crime: experiences of crime and crime control. UNICRI, Rome, pp 289–303
Killias M, van Kesteren J, Rindlisbacher M (2001) Guns, violent crime, and suicide in 21 countries. Can J Criminol 43:429–448
Kleck G (1979) Capital punishment, gun ownership, and homicide. Am J Sociol 84:882–910
Kleck G (1984) The relationship between gun ownership levels and rates of violence in the United States. In: Kates DB Jr. (ed) Firearms and violence: issues of public policy. Ballinger, Cambridge, MA
Kleck G (1988) Crime control through the private use of armed force. Soc Probl 35:1–21
Kleck G (1997) Targeting guns: firearms and their control. Aldine, NY
Kleck G (2004) Measures of gun ownership levels for macro-level crime and violence research. J Res Crime Delinquency 41(1):3–36
Kleck G, DeLone M (1993) Victim resistance and offender weapon effects in robbery. J Quant Criminol 9:55–82
Kleck G, Hogan M (1999) A national case-control study of homicide offending and gun ownership. Soc Probl 46(2):275–293
Kleck G, Kates DB (2001) Armed: new perspectives on gun control. Prometheus, Amherst, NY
Kleck G, McElrath K (1991) The effects of weaponry on human violence. Soc Forces 69:669–692
Kleck G, Patterson EB (1993) The impact of gun control and gun ownership levels on violence rates. J Quant Criminol 9:249–288
Kovandzic T, Vieraitis LM, Yeisley MR (1998) The structural covariates of urban homicide. Criminology 36:569–600
Kovandzic T, Schaffer ME, Kleck G (2012) Gun prevalence, homicide rates and causality: a GMM approach to endogeneity bias. In: Gadd D, Karstedt S, Messner SF (eds) The Sage handbook of criminological research methods. Sage, London
Land K, McCall PL, Cohen LE (1990) Structural covariates of homicide rates. Am J Sociol 95:922–963
Lott JR Jr (2000) More guns, less crime, 2nd edn. University of Chicago Press, Chicago
Marvell TB, Moody CE Jr (1991) Age structure and crime rates: the conflicting evidence. J Quant Criminol 7(3):237–273
McDowall D, Loftin C (1983) Collective security and the demand for handguns. Am J Sociol 88:1146–1161
Moody CE, Marvell TB (2005) Guns and crime. South Econ J 71:720–736
Newey WK (1985) Generalized method of moments specification testing. J Econom 29:229–256
Okoro CA, Nelson DE, Mercy JA, Balluz LS, Crosby AE, Mokdad AH (2005) Prevalence of household firearms and firearm-storage practices in the 50 states and the District of Columbia. Pediatrics 116:e370–e376
Rice DC, Hemley DD (2002) The market for new handguns. J Law Econ 45:251–265
Sampson RJ (1986) Crime in cities. In: Reiss AJ Jr, Tonry M (eds) Communities and crime. University of Chicago Press, Chicago, pp 271–311
Schaffer ME (2007) XTIVREG2: Stata module to perform extended IV/2SLS, GMM and AC/HAC, LIML and k-class regression for panel data models. http://ideas.repec.org/c/boc/bocode/s456501.html
Sloan JH, Kellermann AL, Reay DT, Ferris JA, Koepsell T, Rivara FP, Rice C, Gray L, LoGerfo J (1990) Handgun regulations, crime, assaults and homicide. N Engl J Med 319:1256–1262
Southwick L Jr (2000) Self-defense with guns: the consequences. J Criminal Just 28:351–370
Staiger D, Stock JH (1997) Instrumental variables regression with weak instruments. Econometrica 65:557–586
Stock JH, Watson MW (2007) Introduction to econometrics, 2nd edn. Addison-Wesley, Pearson
Stock JH, Yogo M (2005) Testing for weak instruments in linear IV regression. In: Andrews DWK, Stock JH (eds) Identification and inference for econometric models: essays in honor of Thomas Rothenberg. Cambridge University Press, Cambridge, pp 80–108. Working paper version: NBER Technical Working Paper 284, 2002. http://www.nber.org/papers/T0284
Tark J, Kleck G (2004) Resisting crime: the effects of victim action on the outcomes of crimes. Criminology 42(4):861–909
US Bureau of the Census (1990) Census 1990 summary file 3 (SF3)—sample data, table P006 urban and rural. Retrieved 7 February 2005 from US Census http://factfinder.census.gov
US Bureau of the Census (1994) County and City Data Book, 1994. US Government Printing Office, Washington, DC
US Federal Bureau of Investigation (FBI) 1990–2000, 2006. Crime in the United States 1989 [–1999] Uniform crime reports. US Government Printing Office, Washington, DC
US National Center for Health Statistics (1997) Limited access versions of mortality detail files, 1987–1993, with location detail, supplied to third author. US Department of Health and Human Services, Hyattsville, MD
Vieraitis LM (2000) Income inequality, poverty, and violent crime: a review of the empirical evidence. Soc Pathol 6(1):24–45
White H (1994) Estimation, inference and specification analysis. Cambridge University Press, Cambridge
Wright JD, Rossi PH (1986) Armed and considered dangerous. Aldine, New York
Zimring FE, Hawkins G (1997) Crime is not the problem. NY, Oxford
Acknowledgments
Seminar audiences at Aberdeen, Bologna, Dundee, Edinburgh, IZA (Bonn), Moscow, Strathclyde and Zurich provided useful comments and suggestions on earlier versions of this paper. We are grateful to John DiNardo, David Drukker, John Earle, Andrea Ichino, Austin Nichols, Steven Stillman, Jeff Wooldridge, the editor and two anonymous referees for helpful comments, discussions and suggestions. The usual caveat applies.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Data and Sources
We use cross-sectional data for US counties which had a population of 25,000 or greater in 1990, and for which relevant data were available (N = 1,456). Alaska and Washington, DC were excluded from the analysis: the former, because we did not have compatible data for one of our instruments (voting in 1988); the latter, because it is itself a single county and thus drops out of a fixed-effects specification. Data for most county level variables were obtained from the US Bureau of the Census, County and City Data Book, 1994. Other data sources were as follows:
Homicide rates are averages for the 7 years 1987–1993 (bracketing the decennial census year of 1990). Data for each county were obtained using special Mortality Detail File computer tapes (not the public use tapes) made available by the National Center for Health Statistics (US NCHS 1997). The data include all intentional homicides in the county with the exception of those due to legal intervention (e.g., killings by police and executions).
Similar to homicide, data for the percent of suicides committed with guns are also 1987–93 averages and were obtained using special Part III Mortality Detail File computer tapes made available by the NCHS. Unlike widely available public use versions, the tapes permit the aggregation of death counts for even the smallest counties (US NCHS 1997).
Subscriptions per 100,000 county population to three of the most popular outdoor/sport magazines (Field and Stream, Outdoor Life, and Sports Afield) in 1993 were obtained from Audit Bureau of Circulations (1993). In the earlier version of this paper, we used a principal components index based on the three separate subscription rates; the measure we use here is more convenient and generates almost identical results.
The percent of the county population voting for the Republican candidate in the 1988 Presidential election is from ICPSR (1995). Rurality measures are from US Bureau of the Census (1990).
The statistical package Stata was used for all estimations. The main IV/GMM estimation programs, ivreg2 and xtivreg2, were co-authored by one of us (Schaffer), and can be freely downloaded via the software database of RePEc.Footnote 31 For further discussion of how the estimators and tests are implemented, see Baum et al. (2003, 2007, 2008), Schaffer (2007), and the references therein.
Appendix 2: The OLS and IV Estimators with Population Heterogeneity
Model Setup
The “true model” is one with population heterogeneity (Eq. 7c in the main text):
Criminal and noncriminal gun prevalence are not separately observable. A proxy for aggregate gun prevalence is available (Eq. 14 in the text):
A single instrument Z i is available that is correlated with both criminal and noncriminal gun prevalence, but the strength of the correlation may differ (Eqs. 11 and 12 in the text):
We assume that \( \pi_{ 1}^{\text{C}} ,\pi_{ 1}^{\text{NC}} \ge 0 \) and at least one is strictly greater than zero, and similarly for \( \delta_{ 1}^{\text{C}} \) and \( \delta_{ 1}^{\text{NC}} \). If gun prevalence is directly observable, the estimating equation is (Eq. 1 in the text):
If only the proxy for gun prevalence is observable, the estimating equation is (Eq. 3 in the text):
The derivations below follow the format of those in Stock and Watson (2007), Appendix 13.4.
The Average Treatment Effect (ATE) of Gun Prevalence
Rewrite Eq. (19) as a “random coefficient” model:
where
and by definition \( g_{\text{i}}^{{}} \equiv g_{\text{i}}^{\text{C}} + g_{\text{i}}^{\text{NC}} \), i.e., our measures of gun prevalence are in levels. The average treatment effect of gun prevalence βATE is:
where μC ≡ E(\( g_{\text{i}}^{\text{C}} \)) and μNC ≡ E(\( g_{\text{i}}^{\text{NC}} \)) and where we make use of the result in Frishman (1971) for the expectation of a ratio; the covariance terms account for the fact that the expectation of the ratio of two random variables does not, in general, equal the ratio of the expectations.Footnote 32 In the special case that total gun prevalence is uncorrelated with the criminal/noncriminal share of gun prevalence, the covariance terms in (27) are zero and the ATE takes the following simple form:
Equations (27) and (28) are Eqs. (8) and (9) in the main text.
OLS Estimation
We consider first estimation of Eq. (23), when total gun prevalence is directly observable. The OLS estimator is
where s denotes a sample covariance and \( \mathop \to \limits^{\text{P}} \) denotes convergence in probability. The numerator is
The denominator is simply
and therefore
The OLS estimator differs from the ATE Eq. (27) for two reasons. First, if criminal or noncriminal guns are endogenous, then the third term in Eq. (32) is nonzero. Second, even if gun prevalence is exogenous and the third term in Eq. (32) drops out, the resulting OLS estimator is a weighted average of \( {{\upbeta}}_{ 1}^{\text{C}} \) and \( {{\upbeta}}_{ 1}^{\text{NC}} \), but the weights differ from those for the ATE in Eq. (27); whereas the ATE weights are relative gun prevalence (plus the Frishman correction for the expectation of a ratio), the OLS weights are driven by the variances and covariances of gun prevalence, i.e., by gun variability. The intuition is that the identifying variation in the estimation of Eq. (23) comes from the variation in criminal and noncriminal gun prevalence, and these may differ. To take an extreme example, if criminal and noncriminal gun prevalence are uncorrelated so that \( \text{cov} (g_{i}^{C} ,g_{i}^{NC} ) \) = 0, and noncriminal gun prevalence varies little or not at all across localities so that \( \text{var} (g_{i}^{NC} ) \)≈0, then the OLS estimator \( \hat{\beta }_{1}^{\text{OLS}} \)will be approximately equal to the impact of criminal guns \( {{\upbeta}}_{ 1}^{\text{C}} \), because the identifying variation in the data is driven solely by variation in criminal gun prevalence.
Next we consider OLS estimation of Eq. (24), when only a proxy is available. To simplify the algebra, we assume that the homicide error u i and the proxy error ν i are uncorrelated with gun prevalence and with each other. The OLS estimator is
The numerator is
since we’ve assumed that the error terms are uncorrelated with gun levels. The denominator is
Thus
Equation (36) shows the OLS estimator using a proxy for gun levels is a weighted average of the criminal and noncriminal effects. The weights sum to less than one because of the var(ν i ) term; this is the attenuation bias attributable to the measurement error in the proxy. The weights on \( {{\upbeta}}_{ 1}^{\text{C}} \) and \( {{\upbeta}}_{ 1}^{\text{NC}} \) now depend not only on gun variability, but also on the relative strength of the correlations between the proxy and criminal/noncriminal gun levels: if \( \delta_{ 1}^{\text{NC}} \) ≫ \( \delta_{ 1}^{\text{C}} \), then the OLS estimator will put a high weight on the noncriminal impact gun prevalence, and vice versa if \( \delta_{ 1}^{\text{NC}} \) ≪ \( \delta_{ 1}^{\text{C}} \). Note that even if \( \delta_{ 1}^{\text{C}} \) = 0, the weight on \( {{\upbeta}}_{ 1}^{\text{C}} \) may be positive if criminal and noncriminal gun prevalence are correlated. Note also that sign{\( \hat{b}_{1}^{\text{OLS}} \)} is not in general a consistent estimator of sign{\( \hat{\beta }_{1}^{\text{OLS}} \)}.
IV Estimation
Again we start with the case where gun levels are observable. The IV estimator can be written
Taking the numerator first,
since Z is exogenous and orthogonal to the error term. Using Eqs. (21) and (22), we have
since Z is also uncorrelated with η. Substituting (39) into (38), we have
Now taking the denominator of (37),
where we have made use of (39). Substituting (40) and (41) into (37), we obtain
which is Eq. (13) in the text, the expression for the LATE estimator when gun levels are observable. The LATE estimator is a weighted average of \( {{\upbeta}}_{ 1}^{\text{C}} \) and \( {{\upbeta}}_{ 1}^{\text{NC}} \), but now the weights are the relative strengths of the correlations between the instrument Z and criminal/noncriminal gun prevalence. Note that, unlike the OLS estimator, the variation in gun prevalence does not affect the weights.
Now consider the case where gun levels are not observable and the IV estimator uses the proxy p:
The numerator is the same as in (37) above. The denominator is
Substituting (40) and (44) into (43) yields
which is Eq. (15) in the text, the expression for the LATE estimator when gun prevalence is proxied by p. Note that this is a scaling parameter (assumed positive) times the probability limit of \( \hat{\beta }_{1}^{\text{IV}} \) given in Eq. (42). Thus sign {\( \widehat{\text{b}}_{1}^{\text{IV}} \)} is a consistent estimator of sign {\( \hat{\beta }_{1}^{\text{IV}} \)}, irrespective of the strength of the correlation between the proxy and criminal/noncriminal gun prevalence.
Appendix 3: Robustness Checks
Checks Using 1990 Data
We tested the robustness of our 1990 results by varying the main specification in a number of ways:
-
1.
Weights. We estimated the main specification but weighted by county population in 1990. Because the results using population weights were sometimes sensitive to the inclusion/exclusion of a small number of large counties, we estimated using both the sample of counties with populations in excess of 25,000 as in the results discussed in the main text, and using a subset of this sample that excludes the roughly 100 counties with populations greater than 500,000.
-
2.
Functional form. We varied the functional form of the estimating equation by using homicide rates in logs and in levels, and by using PSG in logs and in levels.
-
3.
Lagged dependent variable (LDV). One of the anonymous referees suggested we include a lagged measure of the gun homicide rate in the equations to mitigate possible problems of unobserved heterogeneity, i.e., historical factors besides heterogeneity in the criminal population. We note, however, that although this is a useful robustness check, the results of such an estimation are not easily interpreted. In our preferred LATE model, the IV estimator has a very clear interpretation in the presence of heterogeneity in criminality: it is a weighted average of the effects of criminal and non-criminal gun prevalence. Including lagged homicide as a regressor eliminates this clear interpretation offered by our LATE model.
-
4.
Criminal justice (CJ) controls. Another referee suggested our specifications may suffer from a specific form of omitted variable bias, namely failing to include controls for formal deterrence measures such as police manpower, incarceration rates, or arrest rates. We examine this possibility by re-estimating our model including as controls two of the most widely used measures in the macro-level deterrence literature and for which data for all US counties are readily available: ICPSR county-level data on sworn police officers per capita in 1992Footnote 33 (as a measure of police manpower levels) and a measure of the rate of solving crimes constructed as the ratio of arrests for violent crimes 1989–91 to reports of violent crimes 1989–91.Footnote 34 We use both controls in log form. Unfortunately, incarceration data are not available for all US counties We should note, however, that to the extent that incarceration levels and other omitted criminal justice measures operate at the state-level to reduce county-level rates of homicide, these effects would be captured by our inclusion of state fixed effects.
We re-estimated the main equation using various combinations of the above specifications. The results for gun homicide are reported below in Table 10; the results for nongun homicide are in Table 11. The first row in each table corresponds to the specification discussed in the main text.
GMUR is the gun murder rate; NGMUR is the non-gun murder rate; PSG is defined as in the paper. “LDV” and “CJ” indicate whether a lagged dependent variable or criminal justice measures are included as regressors. In the LDV specifications, the table reports the long-run coefficient on PSG, equal to the coefficient on PSG * 1/(1 − α) where α is the coefficient on lagged homicide. The magnitude of the long-run coefficient can therefore be compared directly to the coefficient on PSG when the LDV is omitted. The “Wt” column indicates whether or not the results weight by 1990 population. The total sample includes counties with a population of at least 25,000 persons in 1990; the “U lim” column indicates whether a subset of counties with a population upper limit of 500,000 persons is used. The HOLS column reports the coefficient on the gun prevalence proxy when it is treated as exogenous; the GMM2S column is the coefficient when treated as endogenous; 2-step efficient GMM is used in both cases. The “F 1st St” column reports the first-stage F statistic; J is the J overidentification statistic; p(J) is the corresponding p value; and N is the sample size. Stars are as in the paper (1, 5, 10 %). Tests are robust to heteroskedasticity and clustering.
The various specifications show that the main results reported in the paper are indeed robust. In the gun homicide estimations, when gun prevalence is treated as exogenous, the estimated impact on gun homicide is generally positive and statistically significant; when it is treated as endogenous, the impact on gun homicide is significantly negative or null. For nongun homicide, the impact of gun prevalence is generally null, whether or not gun prevalence is treated as exogenous or endogenous. The instrument relevance tests are generally satisfactory (a first-stage F statistic in excess of 10), as are the tests of instrument orthogonality (insignificant J statistics).
The different functional forms generate broadly similar qualitative results, but the specifications in which the homicide rate is in logs, as in the results discussed in the main text, tend to generate smaller quantitative impacts than those in which the homicide rate is used in levels. For the reasons discussed in the text, we regard the log specification, and hence the corresponding quantitative results, as preferable. We discuss here, for illustration, the calibrations for the main specifications for gun homicide (full sample, unweighted, no criminal justice controls or LDV) corresponding to a one percentage point increase in PSG. (1) The specification in line 1 of Table 10 is the specification discussed in the main text; the coefficient of −2.41 on PSG/100 implies that a one percentage point increase in PSG would reduce the gun homicide rate by 0.01 times 2.41 or about 2.4 %. (2) When both the gun homicide rate and PSG are in logs, the estimated coefficient on log(PSG) in the main specification is −1.16 (line 5). A one percentage point increase in PSG evaluated at the sample mean of PSG (67 %; see Table 5) is equivalent to a 1/67th or 1.5 % increase; 0.015 times 1.16 means a fall in the gun homicide rate of 1.7 %. (3) When both the gun homicide rate and PSG are in levels, the coefficient on PSG/100 in the main specification is −28.09 (line 9). A one percentage point increase in PSG would therefore reduce gun homicide by 0.28 persons per 100,000 population. At the sample average of 4.11 gun homicides per 100,000 persons (Table 5), this is equivalent to 6.8 % fall in the gun homicide rate. (4) When the gun homicide rate is in levels and PSG is in logs, the estimated coefficient on log(PSG) in the main specification is −12.35 (line 13). A one percentage point increase in PSG, equivalent to a 1.5 % increase, implies that gun homicide would fall by 0.015 times 12.35 or 0.19 persons per 100,000, equivalent to a 4.5 % fall in the gun homicide rate evaluated at the sample average of 4.11.
Checks Using 1970 and 1980 Data
We constructed datasets for 1970 and 1980 using Census variables, CDC homicide and suicide data, and ICPSR election data. The only variable used in the main specification we did not have available for these earlier years is outdoor sports magazine subscriptions (OMAG), and hence only the voting and veterans variables were available as instruments.
The gun and nongun homicide results for 1970 are reported in Tables 12 and 13, respectively. Because of data availability constraints, some of the control variables are defined slightly differently from the 1990 sample. Age structure categories are 0–14, 15–19, 20–24, 25–44, 45–64, and 65 + . The cutoff for low-income households is $6,000. Inequality is based on household income cutoffs of $6,000 and $25,000. “Hispanic” is based on use of Spanish language. Homicide and PSG are based on 5-year averages, 1968–72. The voting instrument is the percentage voting Republican in the 1968 presidential election. Because 1960 CDC homicide data were not available to us, we do not report the robustness check using a lagged dependent variable. For comparability with the 1990 county coverage, we also report the results of limiting the sample of counties based on 1990 as well as on 1970 population.
The 1970 results for gun homicide are similar to those for 1990: when PSG is treated as exogenous, it has a positive and significant coefficient, but when it is treated as endogenous, the significance disappears or (in a few cases) the coefficient becomes negative and significant. The nongun homicide results are also similar to our 1990 results: PSG has a null or (in a few cases) a negative impact on nongun homicide whether treated as exogenous or endogenous. The overidentification statistics are satisfactory. However, the instruments are weak to very weak, and hence the results should be treated with some caution.
Robustness checks using 1980 data are reported below in Tables 14 and 15. Data definitions are the same as for the 1990 sample, except that the cutoff for low-income households is $8,000, and inequality is based on household income cutoffs of $8,000 and $50,000. Homicide and PSG are based on 5-year averages, 1978–82. The voting instrument is the percentage voting Republican in the 1980 presidential election. Lagged homicide is based on the 1968–72 5-year average. For comparability with the 1990 county coverage, we report the results of limiting the sample of counties based on 1990 as well as on 1980 population. The LDV specifications report the long-run impact of gun prevalence, calculated as noted above.
The results for both gun and nongun homicide in 1980 are similar to those for 1990 and 1970. As with the 1990 data, including a lagged dependent variable does not noticeably change the results. The main difference with the results in term of the specification tests are that J statistic is sometimes high enough to reject at the 5 % level, and that the instruments are weak less often. We note that when the first-stage F statistic is satisfactorily high (near or above 10), the J statistic is also satisfactorily low, and these particular estimations are consistent with our 1990 results (i.e., PSG typically has a negative and significant coefficient in the gun homicide estimations). Again, however, because of the weakness of the instruments, the results should be treated with some caution.
Rights and permissions
About this article
Cite this article
Kovandzic, T., Schaffer, M.E. & Kleck, G. Estimating the Causal Effect of Gun Prevalence on Homicide Rates: A Local Average Treatment Effect Approach. J Quant Criminol 29, 477–541 (2013). https://doi.org/10.1007/s10940-012-9185-7
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10940-012-9185-7