Crime, gun control, and the BATF: the political economy of law enforcement.

INTRODUCTION

The Bureau of Alcohol, Tobacco and Firearms ("BATF"), an agency of the United States Treasury Department, has wide-ranging law enforcement responsibilities, which include the investigation of crimes involving guns, explosives, and illicit drugs. (1) Its execution of these duties has been the subject of considerable controversy. (2) The Bureau's participation in incidents at Ruby Ridge, Idaho, and Waco, Texas, resulting in over one hundred deaths (including that of one federal agent), triggered serious questioning of the competence of the BATF's leadership, (3) as well as its methodology for establishing law enforcement priorities. (4)

The latter issue has been a longstanding point of contention between the BATF and its critics. According to United States Representative John H. Dingell (D-MI), for example, "The goal of the agency appears to be less the prosecution of criminals and persons unlawfully engaged in the illegal use of firearms than in the manufacturing of a statistical record of persons who have committed some technical violation of the 1968 Gun Control Act." (5) A February 1982 report of the Senate Judiciary Committee similarly concluded, "approximately 75 percent of BATF gun prosecutions were aimed at ordinary citizens who had neither criminal intent nor knowledge, but were enticed by agents into unknowing technical violations." (6) Wayne LaPierre summarizes these criticisms in the following terms: "Charged with enforcing federal gun control laws, federal agents persecute and entrap citizens who have done nothing wrong and would never contemplate doing anything wrong." (7)

This Study explores whether conclusive evidence exists to prove that the BATF systematically harasses responsible gun owners. Using cross-sectional data from the fifty states and the District of Columbia for the year 1995, this Study finds, other things equal, BATF agents tend to refer more cases for criminal prosecution to United States Attorneys in states where more citizens belong to the National Rifle Association ("NRA"). (8) Evidence that violent crime rates are lower in states with large numbers of NRA members tends to prove this positive relationship between criminal referrals and NRA membership signifies harassment, rather than cost-effectiveness, in the allocation of scarce law enforcement resources. Moreover, U.S. Attorneys tend to decline to prosecute more of the cases referred to them by BATF agents in high NRA membership states. (9)

This Study also sheds light on the non-existent impact of various state gun control laws on violent crime rates, and on the crime-deterring potential of laws allowing private citizens to carry handguns concealed about their persons. Consistent with the work of Lott and Lott and Mustard, violent crime rates are significantly lower in states where, under general conditions, local police officials "shall issue" concealed-carry permits to adults, (10) except those with prior criminal records or histories of mental illness. (11) The chief contribution of this Study, though, is to add a new dimension to interest-group theories of regulation, namely that regulatory agencies like the BATF may use their discretionary law enforcement authority selectively, not only to channel benefits to special-interest groups in return for political support, (12) but also to quash dissent by harassing members of organizations that oppose the agency's mandate, or are critical of its methods.

Part I of this Study explores some of the historical background and institutional details surrounding the Bureau of Alcohol, Tobacco, and Firearms. Part II presents the data and empirical results used in and stemming out of this Essay's statistical study. Finally, this Study ends with some concluding remarks.

I. A CONCISE HISTORY OF THE BATF

Because of its unique history, the mandate of the Bureau of Alcohol, Tobacco, and Firearms covers a broad range of seemingly unrelated law enforcement responsibilities. (13) The agency's activities run the gamut from revenue collection to crime prevention; it pursues rumrunners, cigarette smugglers, and international terrorists. (14)

Federal excise taxes have been levied on distilled spirits since the Republic's earliest days, being greeted occasionally by opposition. (15) For example, dissent amongst farmers in western Pennsylvania, for whom it was cheaper to transport liquor than raw grain to markets in the east, erupted into the famous Whiskey Rebellion of 1794. Due to this unpopularity, the federal government needed an enforcement mechanism to ensure taxpayer compliance. (16) A small group of agents comprising the Internal Revenue Service's ("IRS") Alcohol Tax Unit ("ATU") performed these duties until the onset of Prohibition. (17)

Franklin Delano Roosevelt campaigned for the presidency in 1932 on a platform calling for the repeal of Prohibition. (18) His support for reverting to a policy of legal alcohol sales was based in large part on a desire to replace the income tax revenue lost to the sharp decline in economic activity following the stock market crash of 1929. (19) Shortly after Prohibition ended in 1933, the Federal Alcohol Administration Act established license and permit requirements for liquor retailers. (20) The responsibility for enforcing this legislation was delegated to the Federal Alcohol Administration, which was housed administratively within the Treasury Department. (21) The Federal Alcohol Administration was merged with the IRS's Alcohol Tax Unit in 1940, and the new organization, still known as the Alcohol Tax Unit ("ATU"), remained part of the IRS. (22) The ATU became responsible for enforcing the collection of tobacco taxes in 1952; it consequently was re-christened the Alcohol and Tobacco Tax Division. (23)

Even before then, however, the agency's regulatory responsibilities had been expanded markedly by enactment of federal gun control legislation. Enforcement of the 1934 National Firearm Act fell under the ATU's aegis. (24) That law, passed in response to the gang-sterism and violence that had erupted during Prohibition, restricted the sale and possession of machine guns, sawed-off shotguns, and similar weapons. (25) Soon after that legislation, the Federal Firearms Act of 1938 made it a federal crime for convicted felons to procure or transport firearms in interstate commerce. (26)

The Gun Control Act of 1968 imposed stricter and more wide-ranging regulatory controls on the firearms industry. (27) Passed in the wake of the assassinations of President John F. Kennedy and Dr. Martin Luther King, Jr., the law mandated that every manufacturer of firearms and every gun dealer obtain a federal license, keep proper records of transactions, and be subject to periodic inspections. (28) Enforcement responsibility was delegated to the IRS's Alcohol and Tobacco Tax Division. (29) In addition, the 1968 Gun Control Act gave the Division jurisdiction over the criminal possession and use of explosives. (30)

The Division's name was changed to the Alcohol, Tobacco and Firearms ("ATF") Division of the IRS in 1970. (31) The ATF was accorded independent bureau status within the Treasury Department in 1972. (32)

The BATF's gun-control responsibilities were broadened further by the enactment of the so-called Brady Bill in 1993. (33) The law established a mandatory five-day waiting period for the purchase of handguns, and charged local law enforcement officials with the responsibility of checking the backgrounds of handgun purchasers prior to delivery. (34)

As a result of the 1968 Gun Control Act and the Brady Bill, firearms violations have become the BATF's chief area of law enforcement activity. (35) Table 1 shows BATF criminal referrals, by program category, to local, state, and federal prosecutors for fiscal year 1995 and the first six months of fiscal year 1996. (36) Of the agency's 9,583 referrals in 1995, for example, 8,612 (nearly ninety percent) of them involved firearms violations. (37)

Like any other government agency, the BATF has seen its budget grow as its regulatory responsibilities have increased. (38) Total BATF expenditures were nearly $129 million in 1978; (39) that figure increased to $409 million in 1995. (40) Expressed in constant 1982-84 dollars, spending climbed by almost thirty-five percent (from $197.5 million to $265.7 million) over this period.

II. DATA AND EMPIRICAL RESULTS

It is no secret that the BATF has been criticized harshly by defenders of the Second Amendment. (41) Acknowledging past problems, the agency implemented a strategic plan in the mid 1990s in an attempt to regain the public's confidence. (42) The plan's purpose was to provide guidance in setting priorities, allocating resources, and evaluating performance. (43)

Numerous stories about the mistreatment of law-abiding citizens by overzealous agents abound, despite this public relations effort. (44) The National Rifle Association, one of the agency's most caustic censors, ran full-page advertisements in the Washington Post and USA Today on March 1, 1995, pointing to the BATF's "tyrannical record of misconduct and abuse of power" and "contempt for civil rights." (45) The ads drew a quick response from Treasury Undersecretary Ronald Noble, who claimed that:

   [W]hile the NRA spends lavishly on ads to fight the ATF, ATF
   agents put their lives on the line to fight crime.... Rather than
   casting about for villains, the NRA should join the American
   public ... in recognizing and praising the men and women who
   risk their lives to protect public safety. (46)

Then-United States Congressman Charles Schumer (D-NY) echoed these sentiments in an open letter to the NRA: "Your ad does not cite a single specific example, nor a single documented source, to support the overblown evils attributed to [the] ATF.... Your ad is not about truth. It is about the extremism of those who control the NRA." (47)

In what follows, this Study explores this controversy in some detail. Using 1995 data on criminal referrals, the Study finds evidence that lends support to the charges of the BATF's critics.

Information relating to the BATF's activities over the years 1992 through 1995 is collected and maintained by the Transactional Records Access Clearinghouse ("TRAC") at Syracuse University. (48) Federal court districts are the units of observation for the data. The information covers all criminal matters referred by the bureau's agents to U.S. Attorneys nationwide. Excluding Guam, Puerto Rico, the Virgin Islands, and the Northern Mariana Islands, there are ninety such geographical court districts. (49) For twenty-six states plus the District of Columbia, the jurisdiction of the federal court is coextensive with the state's boundaries. (50) The remaining twenty-four states house more than one federal court district. (51) In these multi-district states, the district-level data aggregates into state-level observations by simple summation, or, where appropriate, by constructing weighted averages.

The goal of this Study is to explain variations in the BATF's referrals of criminal cases across the fifty states and the District of Columbia. There is considerable disparity in such activities nationwide. (52) Based on comparisons of criminal referrals per capita, David Burnham, TRAC's co-director, suggested that the BATF has allocated more of its law enforcement resources to rural areas than to large cities where illegal guns arguably are a much greater problem. (53) The raw data strongly support this conclusion. On a per capita basis, the largest number of criminal referrals originated in the State of Montana during 1995 (127 cases for a population of 870,000). (54) Western North Carolina had two hundred referrals out of a population of 2.2 million people, and Northern Florida had 128 cases for 1.4 million people that same year. (55) By way of contrast, in 1995 there were 113 BATF referrals out of a population of nearly sixteen million in the Central District of California, including Los Angeles; Northern Illinois, including Chicago, had ninety-eight cases referred out of eight million people; and the Southern District of New York, which includes Manhattan, had 172 referrals for a population base of over four million. (56)

The BATF denies that it pursues crime in small towns more vigorously than in big cities, and offers two explanations for the apparent geographic disparity in referral activity. (57) First, "local law enforcement agencies often find themselves under-manned, and rely on a federal agency like the ATF to investigate crimes." (58) Second, "federal gun laws are often stricter than state laws, particularly in some states in the West." (59) The BATF accordingly makes more referrals for prosecution in these states, particularly some of those which are located in the (rural) West, under stricter federal laws. (60)

These propositions are testable. The Study's empirical analysis proceeds as follows: the Study first estimates a linear regression equation designed to explain cross-state variations in violent crime rates. The Study then reports regression results for criminal referrals, and, to further explore the harassment hypothesis, for prosecutions declined.

A. Violent Crime

The roles played by gun ownership and gun laws in the production function of violent crime are topics of contentious debate. (61) Lott and Lott and Mustard, for example, evoked impassioned reactions by reporting evidence that, other things equal, violent crime rates are lower in states where private citizens easily can secure licenses permitting them to carry concealed weapons. (62) The authors reasoned that when criminals are uncertain whether potential victims are armed, they rationally commit fewer assaults, substituting instead into crimes against property or other illegal activities where the probability of confronting a potentially armed victim is less. (63) In what follows, this result is reproduced on a different data set. Indeed, the evidence from this Study suggests that concealed-carry laws are the only gun control provisions with a statistically significant crime-deterring effect. Moreover, the Study found that crime rates are lower in states having relatively more law-abiding gun owners.

The Study's findings with respect to violent crimes--murder, rape, and (armed) robbery, by and large--are based on estimates of the following linear regression equation for 1995 (64):

VICRIME = [[alpha].sub.0] + [[alpha].sub.1]NRAMEMBERS + [[alpha].sub.2]POLICEXP + [[alpha].sub.3]INVESTIGATE + [[alpha].sub.4]POVERTY + [[alpha].sub.5]UNEMPLOYMENT + [[alpha].sub.6]PPURCHASE + [[alpha].sub.7]PCARRY + [[alpha].sub.9]WAIT + [[alpha].sub.10]CCARRY +[[alpha].sub.11]DC.

Variable definitions, descriptive statistics, and data sources are presented in Table 2.

Violent crime rates are hypothesized to depend on the number of responsible gun owners in a state (NRAMEMBERS), local and state government expenditures for police protection services (POLICEXP), the number of BATF agents available to investigate possible criminal violations (INVESTIGATE), demographic variables (POVERTY and UNEMPLOYMENT), (65) and the restrictiveness of state gun laws (PPURCHASE, PCARRY, WAIT, and CCARRY). DC is a binary variable denoting the District of Columbia, which is known as a high-crime area, and is also the only jurisdiction in the nation to ban the sale of handguns.

NRAMEMBERS serves as a proxy for the number of law-abiding gun owners in a state, and is expected to be negatively related to violent crime rates. This relationship will hold if the members of the National Rifle Association are more responsible than the average citizen in the handling of firearms. Alternatively, criminals may be more hesitant to commit violent crimes in states where greater proportions of the citizenry belong to the NRA, and hence, are more likely to possess firearms with which to defend themselves. (66)

There are two possibilities with respect to the availability of law enforcement resources. One possibility is that more BATF investigators and more local and state spending on police protection deter crime. The other possibility is that both states and the BATF tend to allocate more resources to law enforcement in jurisdictions where there is more criminal activity. The algebraic signs on POLICEXP and INVESTIGATE will depend on which direction of causality dominates.

POVERTY and UNEMPLOYMENT are entered as socioeconomic determinants of crime. More individuals are likely to choose criminal occupations in states where the opportunity cost (in terms of foregone income) is lower. The estimated coefficients on these variables, therefore, is expected to be positive in sign.

Four dummy variables, PPURCHASE, PCARRY, WAIT, and CCARRY, control for the stringency of a state's gun laws. The first three of these variables represent regulations that increase the cost of obtaining and using handguns (virtually no state regulates the ownership of rifles). The estimated coefficients on these variables are expected to be negative to the extent that licensing requirements and waiting periods make it more difficult for criminals to obtain the weapons they use in committing violent crimes. CCARRY, on the other hand, tests whether "shall issue" laws allowing private citizens to carry concealed weapons deter or encourage violent crime. Deterrence will dominate if uncertainty makes criminals hesitant to assault (potentially) armed victims. It has alternatively been argued that the carrying of concealed weapons triggers violence as armed citizens use deadly force to defend themselves. (67)

No a priori prediction is made about the sign of DC. Its coefficient will depend on whether the District of Columbia in fact experiences more violent crime than the nation as a whole, on the average, ceteris paribus.

The regression results, using Halbert White's procedure for obtaining heteroscedasticity-consistent standard errors and co-variances, (68) are reported in Table 3. The three specifications essentially tell the same story. Violent crime rates are significantly lower in states where more citizens belong to the NRA. They are higher where local and state governments allocate more resources to police protection services and where the BATF assigns more agents to criminal investigation duties. These results do not imply that more police protection expenditures and federal agents cause more violent crime. Causality undoubtedly flows in the opposite direction.

The results also suggest that crime rates are higher in states where there is more poverty and unemployment. (69) None of the three dummy variables controlling for the stringency of state gun laws is statistically significant, though. Neither requiring individuals to obtain licenses to purchase or carry handguns, nor compelling individuals to wait prior to taking possession of handguns has apparent value in deterring violent crime. (70) But concealed-carry laws indeed have such an effect; other things being equal, crime rates are significantly lower in those states that permit gun owners to carry weapons concealed about their persons, suggesting that criminals are intimidated by the possibility that their victims are armed. Finally, and perhaps surprisingly, violent crime rates are significantly lower in the District of Columbia than elsewhere, when other factors are held constant.

Overall, the regressions explain between eighty and ninety percent of the variation in violent crime rates across the states. With these results as background, we now turn to an examination of the BATF's criminal referral activities.

B. Criminal Referrals

A key indicator of the BATF's law enforcement activities is the number of cases it refers for prosecution to local U.S. Attorneys. Referrals are not a perfect measure of the extent to which federal agents ferret out law violations because, to quote BATF assistant director Patrick Hynes, "one referral could have 10 defendants." (71) The violations charged are not of equal import in every referral, either. This is a common problem with law enforcement statistics. One antitrust case, for example, can represent a minor infraction committed by a single small firm, or a major criminal price-fixing conspiracy involving multiple defendants. But so long as referrals are positively correlated with the underlying, "true" measure of BATF activity, their determinants will also be positively correlated.

This Study hypothesizes that cross-state variations in BATF referrals are explained by variations in violent crime rates, NRA memberships, BATF and local law enforcement resource availabilities, and state gun laws. These considerations lead to the following linear regression specification:

REFERRALS = [[beta].sub.0] + [[beta].sub.1]VICRIME + [[beta].sub.2]NRAMEMBERS + [[beta].sub.3] + [[beta].sub.4]POLICEXP + [[beta].sub.5]PPURCHASE + [[beta].sub.6]PCARRY + [[beta].sub.7]WAIT + [[beta].sub.8]CCARRY .

As explained in Table 2, REFERRALS is the total number of BATF referrals in a state per 100,000 population; all other variables are defined as before.

VICRIME is included to test whether, given the apparent rural-urban disparity in BATF law enforcement activity, the agency, in fact, allocates its resources in a way that shortchanges high-crime, big-city jurisdictions. A negative sign on VICRIME is expected if this perverse resource-allocation theory holds. If, on the other hand, the BATF allocates more resources to states where violent crime rates are higher, VICRIME will be positively correlated with REFERRALS.

The harassment hypothesis is explored by including NRA membership as a possible determinant of BATF referrals. With violent crime rates held constant, there is no reason to expect referral activity to be significantly related to NRAMEMBERS a priori unless, as asserted by the agency's critics, federal agents tend to harass responsible gun owners by charging them with minor, "technical" law violations. This interpretation is reinforced by the evidence reported above that, other things being equal, violent crime rates are lower in those states where more citizens belong to the National Rifle Association. Hence, a positive sign on NRAMEMBERS is consistent with a law enforcement strategy in which the BATF harasses the members of an interest group that oppose restrictions on gun ownership and, pari passu, the agents who enforce them. (72)

More criminal referrals are expected in those states where there are more BATF investigators assigned to ferret out law violations. Fewer referrals are expected where local and state governments spend more on police protection if, as the agency asserts, its agents commonly pursue investigations that would strain the resources of under-manned local law enforcement agencies. Finally, the four gun law dummy variables (PPURCHASE, PCARRY, WAIT, and CCARRY) are included to test the BATF's contention that its workload is greater in states with weak firearms regulations.

Table 4 reports the regression results. Two versions of the REFERRALS equation are estimated by ordinary least squares. The Study also estimates, by two-stage least squares, a third regression specification where both REFERRALS and VICRIME are treated as being determined endogenously. (73)

The coefficient estimates show a positive and statistically significant relationship between BATF referrals and violent crime. An urban-rural "bias" in the bureau's law enforcement priorities is not apparent when other factors explaining BATF referral activity are held constant. The estimates also provide evidence that the BATF's investigators have positive marginal products--there are more criminal referrals in states where there are more agents in place--and that federal agents do, in fact, serve as substitutes for undermanned local police forces (the estimated coefficient on POLICEXP is negative and significant at the one percent level in all three specifications).

More importantly, other things (including violent crime rates) being equal, criminal referrals are significantly more frequent in those states with larger NRA memberships. Moreover, none of the estimated coefficients on the gun law variables come anywhere close to reaching standard levels of statistical significance. These results dispute the BATF's contention that it has more work to do in states whose gun laws are more liberal than federal laws. They support the hypothesis that the agency's law enforcement philosophy includes special attention for NRA members--a finding that is bolstered by some additional evidence reported below.

C. Prosecutions Declined

As a final test of the hypothesis that the BATF engages in the harassment of responsible gun owners, the Study estimated a regression equation designed to explain cross-state variations in the number of criminal matters referred to U.S. Attorneys that the attorneys decline to prosecute. The Authors' conjecture is that U.S. Attorneys decline to prosecute cases referred to them by BATF agents when the evidence of criminal activity is weak or the violation charged is so minor that the cost of prosecution is less than the expected benefit of obtaining a conviction.

The dependent variable in our regression specification is DECLINED, the number of referrals with prosecution declined in a state per 100,000 population. (74) One new independent variable, the weighted average number of days before a decision to decline to prosecute is reached, TBEFDECLINE, (75) is included on the theory that for bureaucratic reasons U.S. Attorneys are less likely to decline to prosecute a case referred to them the more time they invest in evaluating its merits.

The OLS regression results for this estimation are reported below (absolute values of t-statistics in parentheses): (76)

DECLINED = 1.5613 +.0007VICRIME + .O004NRAMEMBERS
                   (1.36)         (1.99) ([double dagger])
                   -.0034POLICEXP +.0905INVESTIGATE
                   (2.11) ([double (1.61)
                   dagger])
                   -.0011TBEFDECLINE
                   (1.73) ([double dagger])
              ([R.sup.2] = .208; F = 2.36 ([double dagger])). (77)

As expected, U.S. Attorneys decline to prosecute fewer numbers of cases referred to them as greater time elapses before such a decision is reached. Just as the BATF makes fewer criminal referrals in jurisdictions where local and state governments spend more on police protection services, prosecution is declined in fewer of the cases where local police forces have more resources, and, hence, tend to rely less on federal agents to conduct investigations. In addition, there is some evidence that greater numbers of weak cases are referred for prosecution in states where more BATF investigators have been assigned, perhaps indicating that flimsy charges are more likely to be brought in districts where more federal agents compete with one another to show productivity results to their superiors. The estimated coefficient on INVESTIGATE just misses being significant at the ten percent level, and not much should be read into this result.

Finally, and consistent with the harassment hypothesis, more referrals for prosecution are declined, ceteris paribus, in states having more NRA members. Taken together with earlier findings suggesting that BATF agents refer more cases for criminal prosecution to U.S. Attorneys in such states--despite the fact that violent crime rates are lower in states with large NRA memberships--the conclusion that federal agents harass responsible gun owners is almost inescapable.

CONCLUSION

This Study reports evidence that the law enforcement activities of the Bureau of Alcohol, Tobacco, and Firearms are influenced by both public-spirited and bureaucratic motives. While the BATF makes more referrals for criminal prosecution to U.S. Attorneys in states where there is more violent crime and where local and state law enforcement agencies have fewer resources to combat it, evidence is also found that the BATF harasses law-abiding gun owners. In particular, more criminal referrals are made in states where the National Rifle Association enrolls relatively large numbers of members, even though violent crime rates are significantly lower in those states. Evidence that the BATF engages in harassment is reinforced by the finding that, other things being equal, U.S. Attorneys tend to decline to prosecute more of the cases referred to them in states having more NRA members. Finally, the analysis provides no support for the arguments that strict gun control laws deter violent crime or that the licensing of private citizens to carry concealed weapons encourages it. Indeed, violent crime rates are significantly lower in those states that allow private citizens to carry concealed weapons, all else equal.

While abuses of the BATF and other law enforcement agencies often produce sensational headlines, scholars seem reluctant to attribute private-interest motives to the policing and judicial institutions of government. This tendency to treat the forces of law and order as being somehow above the fray of ordinary politics is particularly noteworthy in the area of antitrust, where public-interest models of bureaucratic behavior hold sway, despite the accumulation of an empirical literature suggesting that these laws are driven by less laudable ideals. (78) The evidence presented in this Study contributes to a rethinking of the public-interest paradigm. To be sure, the BATF allocates more of its resources to high-crime states and helps fill some of the law enforcement gaps left open by underfunded local police forces. At the same time, however, the bureau apparently targets responsible gun owners, perhaps because these individuals are often the BATF's most vocal critics. Measured both in terms of the lives lost in places like Ruby Ridge, Idaho, and the violent crimes that go undeterred while the agency harasses law-abiding citizens, the opportunity cost of such a law enforcement strategy can be quite high.

APPENDIX

TABLE 1

BATF REFERRALS BY PROGRAM AREA

Program Area    Fiscal Year 1995    Fiscal Year 1996 (a)

  Firearms           8,612                 3,453
 Explosives            409                   177
   Arson               438                   179
  Tobacco               50                    23
  Alcohol               28                    17
   Other                46                    23
   TOTAL             9,583                 3,872

(a) First six months of fiscal year.

Source: Transactional Records Access Clearinghouse, Syracuse
University, at http://trac.syr.edu (last visited Jan. 15, 2003.)

TABLE 2

VARIABLE DEFINITIONS, MEANS, AND STANDARD DEVIATIONS

                                                            Standard
    Variable               Definition              Mean     Deviation

 REFERRALS (1)      Number of BATF referrals         3.29       2.19
                   for prosecution per 100,000
                           population
  VICRIME (2)        Violent crime rate, per       595.69     400.97
                      100,000 persons, 1994
INVESTIGATE (3)        Total BATF criminal           1.15       3.49
                    investigators per 100,000
                           population
  POLICEXP (4)     State and local government     $291.86    $165.03
                    police expenditures, per
                          capita, 1992
 NRAMEMBERS (5)      NRA members per 100,000     1,455.18     692.74
                           population
  POVERTY (6)       Percent of persons below        13.23       4.05
                   poverty level, 1994 (percent)
UNEMPLOYMENT (7)     State unemployment rate         5.26       1.28
                            (percent)
 PPURCHASE (8)     Binary variable indicating         .24        .43
                   states that require a permit
                      to purchase a handgun
   PCARRY (9)      Binary variable indicating         .55        .50
                   states that require a permit
                       to carry a handgun
   WAIT (10)       Binary variable indicating         .43        .50
                   states that impose a waiting
                   period prior to purchasing a
                             handgun
  CCARRY (11)      Binary variable indicating         .61        .49
                     states that permit the
                      carrying of concealed
                             weapons

Note: All data are for 1995 unless otherwise indicated.

(1.) Transactional Records Access Clearinghouse, Syracuse, New York, at
http://trac.syr.edu/ (last visited Jan. 15, 2003)

(2.) ALMANAC OF THE 50 STATES: BASIC DATA WITH COMPRATIVE TABLES 436
(Edith R. Hornor ed., 1996).

(3.) Trasactional Records Access Clearinghouse, supra note 1.

(4.) United States Dep't of Commerce, Bureau of the Census,
Statistical Abstract of the United States 1995, 201 (1995).

(5.) Personal Correspondence from the National Rifle Association, to
Jim F. Couch (on file with author).

(6.) United States Dep't of Commerce, Bureau of the Census, Statistical
Abstract of the United States 1996, 397 (1996).

(7.) Id. at 474.

(8.) National Rifle Association, Index of Gun Laws, available at
http://www.nra.org/gun-laws (last visited Jan. 15, 2003).

(9.) Id.

(10.) Id.

(11.) Id.

TABLE 3

REGRESSION RESULTS (DEPENDENT VARIABLE: VICRIME)

Independent Variables           (1)                   (2)

      Intercept                268.9952             -324.3114
     NRAMEMBERS                  -.1506                -.0937
                                (3.80)                (3.05)
                         ([dagger] [dagger])   ([dagger] [dagger])
      POLICEXP                   1.5439                1.6807
                                (5.28)                (7.10)
                         ([dagger] [dagger])   ([dagger] [dagger])
     INVESTIGATE               228.6481              188.0786
                                (2.58)                (4.79)
                             ([dagger])        ([dagger] [dagger])
       POVERTY                                        34.7299
                                                      (6.77)
                                               ([dagger] [dagger])
    UNEMPLOYMENT
      PPURCHASE                -55.9588                9.1825
                                 (.69)                 (.16)
       PCARRY                   31.1609               22.7249
                                 (.52)                 (.50)
        WAIT                     3.6342               40.7145
                                 (.05)                 (.73)
       CCARRY                 -113.3515              -82.3374
                                (1.82)                (1.69)
                         ([double dagger])     ([double dagger])
         DC                  -5296.163             -4685.267
                                (2.46)                (4.75)
                             ([dagger])        ([dagger] [dagger])
      [R.sup.2]                   .796                  .891
     F-statistic                20.47                 37.17
                         ([dagger] [dagger])   ([dagger] [dagger])

Independent Variables           (3)

      Intercept                 68.9145
     NRAMEMBERS                  -.1434
                                (3.87)
                         ([dagger] [dagger])
      POLICEXP                   1.3367
                                (4.23)
                         ([dagger] [dagger])
     INVESTIGATE               226.6453
                                (3.00)
                         ([dagger] [dagger])
       POVERTY
    UNEMPLOYMENT                50.0672
                                (2.00)
                             ([dagger])
      PPURCHASE                -24.5194
                                 (.28)
       PCARRY                   -2.4950
                                 (.04)
        WAIT                    18.1501
                                 (.26)
       CCARRY                 -124.9448
                                (2.17)
                             ([dagger])
         DC                  -5250.467
                                (2.89)
                         ([dagger] [dagger])
      [R.sup.2]                   .811
     F-statistic                19.58
                         ([dagger] [dagger])

Note: Absolute values of t-statistics in parentheses.

See Table 2 for variable definitions.

([dagger] [dagger]) Significant at the one percent level.

([dagger]) Significant at the five percent level.

([double dagger]) Significant at the ten percent level.

TABLE 4

REGRESSION RESULTS (DEPENDENT VARIABLE: REFERRALS)

Independent            (1)                    (2)
 Variables             OLS                    OLS

 Intercept           2.0265                 2.2882

  VICRIME             .0026                  .0028
                    (1.88)                 (1.93)
                ([double dagger])      ([double dagger])
NRAMEMBERS            .0017                  .0015
                    (3.38)                 (2.82)
               ([dagger] [dagger])    ([dagger] [dagger])
INVESTIGATE           .2499                  .2911
                    (1.69)                 (1.87)
                ([double dagger])      ([double dagger])
 POLICEXP            -.0103                 -.0113
                    (2.85)                 (2.91)
               ([dagger] [dagger])    ([dagger] [dagger])
 PPURCHASE                                   .0908
                                            (.12)
  PCARRY                                     .2063
                                            (.32)
   WAIT                                     -.8261
                                           (1.28)
  CCARRY                                     .5455
                                            (.84)
 [R.sup.2]            .242                   .280
F-statistic          3.66                   2.04
                   ([dagger])          ([double dagger])

Independent            (3)
 Variables          TSLS (a)

 Intercept           1.8122

  VICRIME             .0040
                    (2.03)
                   ([dagger])
NRAMEMBERS            .0017
                    (2.93)
               ([dagger] [dagger])
INVESTIGATE           .2693
                    (1.70)
                ([double dagger])
 POLICEXP            -.0131
                    (2.98)
               ([dagger] [dagger])
 PPURCHASE            .1671
                     (.21)
  PCARRY              .1481
                     (.23)
   WAIT              -.8408
                    (1.29)
  CCARRY              .6451
                     (.96)
 [R.sup.2]            .268
F-statistic          2.06
                ([double dagger])

(a) The instrumental variales are NRAMEMBERS, INVESTIGATE, POLICEXP,
PPURCHASE, PCARRY, WAIT, CCARRY, POVERTY, AND DC.

Note: Absolute values of t-statistics in parentheses.

See Table 2 for variable definitions.

([dagger] [dagger]) Significant at the one percent level.

([dagger]) Significant at the five percent level.

([double dagger]) Significant at the ten percent level.

(1.) Dismantling of the Bureau of Alcohol, Tobacco & Firearms: Hearing Before the Subcomm. on Crime, House Comm. on the Judiciary, 97th Cong. 3 (1981).

(2.) See David B. Kopel & Paul H. Blackman, No More Wacos: What's Wrong with Federal Law Enforcement and How to Fix It 292-310 (1997).

(3.) Polemicists have characterized the BATF as a "rogue agency." See, e.g., WAYNE LAPIERRE, GUNS, CRIME, AND FREEDOM 177-79 (1994). AS long ago as 1980, even public officials otherwise strongly supportive of expansive governmental authority have disparaged the bureau's methods: "If I were to select a jack-booted group of fascists who were perhaps as large a danger to American society as I could pick today, I would pick the BATF. They are a shame and a disgrace to our country." Id. at 177 (quoting Congressman John H. Dingell).

(4.) Id. at 177-200 (supplying details about the Ruby Ridge and Waco incidents in a chapter entitled "BATF Abuses").

(5.) Id.; see KOPEL & BLACKMAN, supra note 2, at 305.

(6.) The Right to Keep and Bear Arms: Hearing Before the Subcomm. on the Constitution, Senate Comm. on the Judiciary, 97th Cong. 23 (1982).

(7.) LAPIERRE, supra note 3, at 178.

(8.) See discussion infra Part II, Conclusion, & App.

(9.) See discussion infra Part II, Conclusion, & App.

(10.) See JOHN R. LOTT, JR., MORE GUNS, LESS CRIME: UNDERSTANDING CRIME AND GUN-CONTROL LAWS 50-96 (1998); see also John R. Lott, Jr. & David B. Mustard, Crime, Deterrence, and Right-to-Carry Concealed Weapons, 26 J. LEGAL STUDIES 1, 18-26 (1997).

(11.) Federal law prohibits the sale of firearms to these classes. See 18 U.S.C. [section] 922(d) (2000).

(12.) See ROBERT E. MCCORMICK & ROBERT D. TOLLISON, POLITICIANS, LEGISLATIONS, AND THE ECONOMY: AN INQUIRY INTO THE INTEREST-GROUP THEORY OF GOVERNMENT 7-12 (1981) (correlating enforcement activity with house sizes, legislative size, wealth, and population); Sam Peltzman, Toward a More General Theory of Regulation, 19 J.L. & ECON. 211,213-31 (1976); George J. Stigler, The Theory of Economic Regulation, 2 BELL J. ECON. & MGMT. Sci. 3, 3-6 (1971).

(13.) WILLIAM J. VIZZARD, IN THE CROSS FIRE: A POLITICAL HISTORY OF THE BUREAU OF ALCOHOL, TOBACCO, AND FIREARMS 1-16 (1997).

(14.) Id.

(15.) Brenda Yelvington, Excise Taxes in Historical Perspective, in TAXING CHOICE: THE PREDATORY POLITICS OF FISCAL DISCRIMINATION 31, 32-33 (William F. Shughart II ed., 1997).

(16.) Id. at 33-34.

(17.) VIZZARD, supra note 13, at 2.

(18.) Yelvington, supra note 15, at 40.

(19.) See id.

(20.) See Federal Alcohol Administration Act, ch. 814, 49 Stat. 977 (1935) (current version at 27 U.S.C. [subsection] 201-219a (1988)); A HISTORICAL GUIDE TO THE U.S. GOVERNMENT 40 (George T. Kurlan ed., 1998).

(21.) Id. at 40-41.

(22.) Id. at 41.

(23.) Id.

(24.) National Firearms Act, ch. 757, 48 Stat. 1236 (1934) (current version at 26 U.S.C. [subsection] 5801-5872 (1994)).

(25.) Id.

(26.) Federal Firearms Act of 1938, ch. 850, 52 Stat. 1250 (1958) (current version at 18 U.S.C. [subsection] 921-927 (1994)).

(27.) Gun Control Act of 1968, Pub. L. No. 90-618, 82 Stat. 1213 (codified as amended at 18 U.S.C. [subsection] 921-927).

(28.) Id.

(29.) A HISTORICAL GUIDE TO THE U.S. GOVERNMENT, supra note 20, at 40.

(30.) Id.

(31.) Id.

(32.) Id.

(33.) The Brady Handgun Violence Prevention Act, Pub. L. No. 103-159, 107 Stat. 1536 (1993) (codified as amended at 18 U.S.C. [section] 925(a)).

(34.) Id.

(35.) See infra App., Tbl. 1.

(36.) Infra App., Tbl. 1.

(37.) Infra App., Tbl. 1.

(38.) OFFICE OF MGMT. & BUDGET, BUDGET OF THE UNITED STATES GOVERNMENT FOR FISCAL YEAR 1997, at 803 (1996); U.S. DEP'T OF JUSTICE, BUREAU OF ALCOHOL, TOBACCO AND FIREARMS--BUDGET FOR FISCAL YEARS 1975-2002, available at http://www.usdoj.gov/jmd/budgetsummary/btd/1975_2002/btd01batf.htm (last visited Jan. 15, 2003).

(39.) U.S. DEP'T OF JUSTICE, supra note 38.

(40.) OFFICE OF MGMT. & BUDGET, supra note 38, at 803.

(41.) See Letter from Ronald K. Noble, Under Secretary, Department of the Treasury, to Tanya K. Metaska, Executive Director, National Rifle Association of America (Mar. 3, 1995), available at http://elfie.org/-croaker/treasury.html (last visited Jan. 15, 2003); Letter from Charles Schumer, Congressman, House of Representatives, to Tanya K. Metaska, Executive Director, National Rifle Association of America (Mar. 2, 1995), available at http://elfie.org/-croaker/shumer.html (last visited Jan. 15, 2003).

(42.) KOPEL & BLACKMAN, supra note 2, at 307-10.

(43.) Id.

(44.) See, e.g., id. at 308.

(45.) Letter from Ronald K. Noble to Tanya K. Metaska, supra note 41.

(46.) Id.

(47.) Letter from Charles Schumer to Tanya K. Metaska, supra note 41.

(48.) Transactional Records Access Clearinghouse, Syracuse University [hereinafter TRAC], at http://trac.syr.edu (last visited Jan. 15, 2003).

49. See 28 U.S.C. [subsection] 81-133 (2000) (laying out the United States geographical court districts).

(50.) See id. [section] 133.

(51.) See id.

(52.) TRAC, supra note 48, at http://trac.syr.edu/tracatf/findings/aboutATF/atfRegionalVar.html (last visited Jan. 15, 2003).

(53.) Karen Gullo, ATF More Active in Small Towns, TIMES DALLY (Florence, Ala.), Sept. 8, 1996, at 4A.

(54.) TRAC, supra note 48, at http://trac.syr.edu/tracatf/findings/95/criminal/refR95. html (last visited Jan. 15, 2003).

(55.) Gullo, supra note 53, at 4A.

(56.) Id.; see TRAC, supra note 48, at http://trac.syr.edu/tracatf/findings/95/criminal/ refR95.html (last visited Jan. 15, 2003).

(57.) Gullo, supra note 53, at 4A.

(58.) Id.

(59.) Id.

(60.) Id.

(61). See, e.g., David B. Kopel, Peril or Protection? The Risks and Benefits of Handgun Prohibition, 12 ST. Louis U. PUB. L. REV. 285,287-323 (1993) (pointing out various arguments and constitutional issues involved in the debate over handgun prohibition).

(62.) LOTT, supra note 10, at 94-96; Lott & Mustard, supra note 10, at 64.

(63.) LOTT, supra note 10, at 114-15; Lott& Mustard, supra note 10, at 64.

(64.) Data availability issues restrict the empirical analysis to 1995. That is the only year for which information was obtainable on the number of NRA members by state.

(65.) In addition to these two demographic variables, the Study also estimated regressions including the fraction of the population between the ages of five and thirty-four, and the percentage of persons living in urban areas. Neither of these variables proved statistically significant either when entered by themselves, or in specifications including POVERTY or UNEMPLOYMENT.

(66.) NRA members may, in other words, confer a positive externality on their neighbors. Such a possibility is supported by evidence that widespread firearms ownership in the United States is consistent with the significantly lower number of so-called "hot burglaries" (residents are at home when the criminals strike) committed in the United States, compared with countries such as Canada and the United Kingdom, whose laws severely limit private gun ownership. See DAVID B. KOPEL, THE SAMURAI, THE MOUNTIE, AND THE COWBOY 418 (1992).

(67.) See, e.g., Philip J. Cook, The Role of Firearms in Violent Crime, in CRIMINAL VIOLENCE 236 (Marvin E. Wolfgang & Neil A. Werner eds., 1982); Philip J. Cook, The Technology of Personal Violence, 14 CRIME & JUST.: ANN. REV. RES. 1, 57 (1991); Philip J. Cook et al., Regulating Gun Markets, 86 J. CRIM. L. & CRIMINOLOGY 59, 6165 (1995); Franklin Zimring, Is Gun Control Likely to Reduce Violent Killings?, 35 U. CHI. L. REV. 721,721-25 (1968). See also H. RICHARD UVILLER, VIRTUAL JUSTICE 95 (1996). "More handguns lawfully in civilian hands will not reduce deaths from bullets and cannot stop the predators from enforcing their criminal demands and expressing their lethal purposes with the most effective tool they can get their hands on." Id. Private gun ownership, in other words, kindles an arms race that law-abiding civilians cannot hope to win. See id. For a listing of these dummy variables, see App.

(68.) Halbert White, A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity, 48 ECONOMETRICA 817, 821-27 (1980).

(69.) When both of these variables are entered in the same regression, their estimated coefficients remain positive, but only POVERTY is different from zero at standard levels of statistical significance.

(70.) The insignificance of WAIT may be due to the Brady Bill's preemption of state gun laws.

(71.) Gullo, supra note 53, at 4A.

(72.) A positive coefficient on NRAMEMBERS is also consistent with bureaucratic incentives to produce "visible" output easily observed by the agency's oversight committees in Congress. Cotton M. Lindsay, A Theory of Government Enterprise, 84 J. POL. ECON. 1061, 1065 (1976). The BATF can justify requests for more generous operating budgets by pursuing large numbers of minor gun-law violations, rather than devoting its resources to fewer "big" cases. The bureau simply selects easy targets to appear busy and can do so in states where the membership of the NRA is relatively large, gun ownership is more widespread and, hence, BATF agents are afforded more opportunities for uncovering "technical" violations of the law. Although the analysis does not distinguish between these two explanations, given the acrimonious relations between the NRA and the BATF, the harassment hypothesis seems more plausible.

(73.) Strictly speaking, two-stage least squares ("TSLS") are not necessary in this case. On the basis of the simple correlation coefficient between the residuals of the regression results reported in the second columns of Tables 2 and 3 ([r.sup.2] = -.104), the system is recursive, and so the two equations can be estimated separately without introducing simultaneous-equations bias.

(74.) The mean of the dependent variable is 1.29, with a standard deviation of .83.

(75.) In those states encompassing more than one federal court district, TBEFDECLINE is the weighted average of the mean number of days elapsing before a decision to decline prosecution is reached, where the weights are the total number of referrals for prosecution in each district. TBEFDECLINE is the simple state average otherwise. The mean of this variable is 379.88 days, with a standard deviation of 154.57 days.

(76.) ([dagger]) indicates significance at the five percent level of statistical confidence; ([double dagger]) indicates significance at the ten percent level.

(77.) Similar results (not reported here) are obtained using two-stage least squares.

(78.) FRED S. MCCHESNEY & WILLIAM F. SHUGHART II, THE CAUSES AND CONSEQUENCES OF ANTITRUST: THE PUBLIC-CHOICE PERSPECTIVE 231-42 (1995); WILLIAM F. SHUGHART II, ANTITRUST POLICY AND INTEREST-GROUP POLITICS 36-51 (1990).

William F. Shughart II, Professor, Department of Economics, University of Mississippi. The Authors are grateful to Bernie Hoerr, Director of Membership Programs, National Rifle Association, for granting access to the NRA membership data on which this Study analysis is based. John Lott and Hilary Shughart provided useful comments on an earlier version of this Essay; J. G. Chen supplied able research assistance. As is customary, however, the Authors accept full responsibility for any errors.

Jim F. Couch, Professor, Department of Economics, Finance and Quantitative Studies, University of North Alabama.