ONLINE APPENDIX TO FEAR OF OBAMA: AN EMPIRICAL STUDY OF THE DEMAND FOR GUNS AND THE U.S. 2008 PRESIDENTIAL ELECTION ∗ Emilio Depetris-Chauvin
This Online appendix is not intended for publication.
It includes several tables not included
in the main manuscript due to space constraints. It also presents and discusses several robustness checks. Finally, this online appendix explains the methodology for the construction of the evolution of the stock of guns in the U.S.
1
Robustness Check
All the specications presented in Table 1 to 5 in the main manuscript are based on the period March 2007-December 2009. As a robustness check, I exclusively focus on the election race, omitting all state-month observations after November 2008. Column 1 in Table S.5 shows that the Obama eect remains strongly statistically signicant: a 10-point increase in the probability of Obama victory is associated with an increase in the demand of guns of 5.8 percent. The coecients for the main Obama term and its interactions in column 2 are jointly statistically signicantly dierent from zero (Chi squared-Statistic is 198). The implied Obama eect evaluated at the mean values of interacted variables is 0.63. Notably, the point estimates for the interaction between the probability of Obama winning the election and the weakness of state gun control measure substantially increases in size and gains more precision. The results for average prejudice interaction term are statistically weaker. The introduction of month xed eects in column 3 does not alter the key ndings. The sharp drop in sample size in these specications is likely to partially account for the reduction in the precision of the estimation of some of the coecients. The identication strategy of this paper somehow assumes that individuals are informed about the evolution of the each probability implied from IEM data. Ordinary individuals may have difculties understanding the concepts involved in the calculation of those probabilities and their im-
1
plications.
In addition, general knowledge of the existence of IEM data may not be widely spread.
Therefore, it is conceivable that individuals end up using poll data as their primary source of information (or proxy) to update the likelihoods of each of the election outcomes. In Table S.6 I use ∗ Universidad de los Andes - CEDE. Email:
[email protected]
1 Another
potential objection is, in line with Tversky and Kahneman (1974), that individuals may make decisions
based of subjective assessments of probabilities which, in some cases, may be quite dierent from the true (objective) probabilities.
1
normalized poll data for both the 2008 presidential and Democrat primary elections instead of the IEM data to account for the likelihood of Obama and other Democrat winning the election.
2
The
results are similar to and consistent with the ones obtained in my preferred specications in Tables 3 and 4 in the main manuscript. In my next robustness check exercise I focus on alternative measures for prejudice.
All the
specications include state and month xed-eects. Sample sizes across specications vary depending on data availability. In column 1 of Table S.7 I account for the number of racially motivated hate crimes reported to the FBI in 2008 (per million of people covered by the agencies reporting to the FBI). The coecient estimate for its interaction term with the probability of Obama winning the election is positive but statistically insignicant at the standard levels of condence. The accuracy of FBI's hate crime data is subject of debate among experts (see, for example, Perry (2001); and Rubenstein (2004)) since measurement error is likely to be present.
3
For the second specication
in column 2 I use a racial attitude measure from Fisman et al (2008). interaction term is also positive and strongly statistically signicant.
4
The coecient for this
A two standard deviation
increase in the level of prejudice is statistically associated with almost 16 percent increase in the demand for guns when Obama is elected. In column 3 I use Mas and Moretti (2009)'s measure of
5
prejudice.
I nd similar results. In the last specication in column 4 I use the average prejudice
index from Charles and Guryan (2008) and represents an average over a 30 years period for an aggregate measure of racially prejudiced sentiments at the the state level.
2 The
6
Again, I nd similar
source of the poll data is pollster.com. In order to have some measure of comparability between my proxy
and the probability data from IEM I proceed as follows. I construct the normalized (i.e: I transform the shares of all candidates so they add to one) poll data for Obama winning the primary by using the monthly average of the polls asking for the favorite Democrat candidate. Then I follow the same procedure using the polls on Obama vs McCain (assuming that McCain was the Republican candidate with probability equal to 1) and obtaining a proxy for the unconditional probability of Obama winning the election. By taking the product of the two normalized poll gures (i.e: Obama winning the primary and Obama vs McCain) I obtain the poll data on the likelihood of Obama winning the election.
In order to construct the poll proxy for other Democrat candidate winning election I use the same
procedure using the probability of Obama losing the primary and the data on polls Clinton vs McCain (assuming that Clinton was the only Democrat alternative to Obama). Note that, as in previous specications, the value 1 in November 2008 and afterward while
3 Even
P (Democ−Obama )
P (Obama)
takes
takes the value 0 from September 2008 on.
the FBI cautions of making cross-sectional comparison based on hate crime data. Furthermore, according
to the Bureau of Justice Assistance (1997) ...many police jurisdictions, especially those in rural areas, simply do not have the manpower, inclination, or technical expertise to record hate crimes, and other jurisdictions fear that admitting the existence of hate crimes will cause their communities cultural, political, and economic repercussions. Law enforcement and citizen attitude toward reporting hate crimes might be correlated with the true level of prejudice of a given location. Since hate crime under-reporting would be more likely in states where race prejudice is in some degree more institutionalized, this measurement error might be introducing a downward bias in the estimation of the relevant coecient.
4 This
measure accounts for the fraction of respondents that answered yes to the GSS's question Do you think
there should be laws against marriages between (Negroes/Blacks/African-Americans) and whites? They compute this variable at the state level using all the responses for the period 1988-1991.
Note that this variable take into
consideration all the respondents no matter their race so it will not be picking up the eect of the specic negative attitude towards black from white people (as in Charles and Guryan (2008)) but an overall prejudice eect.
5 Mas
and Moretti (2009) use a similar measure as in Fisman et al (2008) since it is based on the response to
same GSS's question. The most important dierence is that Mas and Moretti (2009) take into consideration only the fraction of whites who support anti-interracial-marriage laws.They also expand the sample size including all GSS's waves between 1990 and 2006.
6 Using
multiple waves (1972-2004) of the GSS and responses from white people aged 18 and older to more than 20
dierent racial prejudice questions, Charles and Guryan (2008) build a proxy for prejudice at the state level. They
2
results: a two standard deviation increase in state's prejudice index is statistically associated with a 10-percent increase in the demand for guns when Obama is been elected. Table S.8 includes further robustness checks. All specications include state and month xed effects. Therefore, they are intended to be compared with column 4 in Table 5 in the main manuscript. In column 1 I add the four interactions between the probability of other Democrat winning the election and the relevant state characteristics (i.e: gun control weakness, prejudice, Republican prevalence, and gun ownership rates) to assure that previous results were not just picking up a more general Democrat eect.
7
Only the coecient for the
P (Democ−Obama )t
interaction term with prej-
udice is statistically signicant at the standard levels of condence (results not reported in table). The point estimates for the interaction terms are broadly consistent with my previous ndings. Nonetheless, the coecient estimates for the interaction term for average prejudice become larger and statistically more signicant whereas the weak state gun control interaction term is somewhat smaller (albeit statistically signicant at the 1 percent level). In the specication in column 2 I explore whether any of my main results is driven by the South. According to Stephens-Davidowitz (2014)'s racial animus index, southern states tend to have levels of prejudice above the national average. In addition, southern states also had larger increases in the demand for guns after the election of Obama. Thus, in column 2, I omit from the sample all the
8
southern states. Results remain similar.
From column 3 to 6 I consider state heterogeneity in other confounding variables.
First, in
column 3 I add an interaction term exploiting cross-state variation in public perception about US economy for 2008.
9
Although states in which consumers had lower condence in both the current
and future state of the economy experienced higher increases in the demand for guns when the odds of the election of Obama increased (point estimate not reported in table), results obtained in Table 5 are not aected. In column 4 I use information from Google Trends about the relative intensity of searches including the terms stimulus and economic stimulus for the period January-September
10
2008 as another proxy for public concern's about the state of the economy.
Consistently with the
previous nding, states with relative more search volume for the two terms also experienced a higher focus only on questions that are exclusively related to racial prejudice, omitting in their analysis the ones touching on government policies and race. After normalizing responses, they construct a individual-level prejudice index for each respondent and then aggregate these gures at the state level.
7I
acknowledge that applying my regression analysis to the 1992 election may be arguably a better robustness
check. In fact, federal tax receipt from rearms sales rose by an average 25 percent during the 4 quarters following the election of Bill Clinton and concurring with the discussion in Congress of the Brady Handgun Violence Prevention Act and the Federal Assault Weapons Ban. Unfortunately, there is no monthly data on cross-state variation in gun purchases -or any proxy- for that period (NIC's data is available since 1999).
8 Results
are unaltered when I omit each southern census region at a time (point estimates not shown). None of
the three southern sub-regions appear to be driving the main results.
9I
use Gallup's economic condence index which is based in 2 questions regarding public perception of the current
and future economic conditions.
Residents of states mostly aected by unemployment in 2008 (such as Michigan
and Rhode Island) had the most negative perceptions in 2008. Oil-producing states (such as Texas, Utah and North Dakota) that were somehow beneted from the surge in oil prices were the least negative in 2008 (Saad, 2009).
10 Each
state's score (i.e: Search Volume Index) is relative to the state with the highest score (which is always
100) of relative searches for a particular term. In my specication I use the average of the states' score for the terms stimulus and economic stimulus. Using exclusively either the former or the latter does not aect the results (results not shown)
3
demand for guns when Obama election probability was higher. The inclusion of this new interaction term does not aect the statistical signicance of weak gun control interaction term. In column 5 of Table S.8 I include controls for six lags of monthly crime counts at the state level.
11
While the six lags are jointly signicantly dierent from zero, the previous results are
not substantially aected. In the last column of Table S.8 I add four additional interaction terms to account for state characteristics that may confound with my prejudice and weak gun control measures. I include black population (relative to white population) which is strongly and positively correlated with my measure of prejudice, average crime rate in the period 2006-2007 (positively correlated with weak gun control), (log of ) income per capita in 2007 (negatively correlated with the two measures), and rural population share (positively correlated weak gun control). The coecients for weak gun control is still statically signicant at 1 the percent level after controlling for this new set of interactions (adding one control at a time does not aect the conclusions, point estimates not shown).
The interaction term for the prejudice variable is statistically insignicant at the
standard levels of condence in most of the specications of Table S.8 (p-values between 0.11 and 0.15 depending on the specication). Although the size of the point estimates across specications are similar to the ones in Table 5, their standard errors are larger.
2
Construction of the Stock of Guns for 1994-2012
I exploit data on rearm manufacturing, exporting and importing for non-military use in the United States to construct series of the stock of rearms by type of guns for the time period 1994-2012. These data come from the Bureau of Alcohol, Tobacco, Firearms, and Explosives (U.S. Department of Justice). I assume a very simple law of motion for the stock of guns:
Stocki,t = Stocki,t−1 + Yi,t − Xi,t + Ii,t where
i
spectively.
and
t
denote type of gun (handguns, ries, shotguns, or other longguns) and year, re-
Yi,t , Xi,t
and
Ii,t
12
are the net production, exports, and imports of guns.
Three things
are worth mentioning: First, I am assuming no depreciation in the stock of guns (rearms are extremely durable products). Second, I am not taking into account changes in inventories. Anecdotal evidence appears to indicate that rearms dealers faced large drops in their inventories during the 2008-2009 period.
Finally, data from the Bureau of Alcohol, Tobacco, Firearms, and Explosives
includes production and trade of rearms destined not only to households but also to law enforcements. Nonetheless, law enforcement agencies tend to sell their rearms when needed to be replaced (thus, the overstatement of the ow of guns into private hands should not be that large). I set the initial stock in 1994 based on Cook and Ludwig (1997) estimates which includes only rearms owned
11 I excluded Alabama and Florida due to the lack of monthly crime data. 12 Data can be accessed here https://www.atf.gov/content/About/statistics
4
by households. The following table shows the initial stocks estimated by Cook and Ludwig (1997) and the stock I estimate as of December 2012:
Type of Gun
Cook and Ludwig Estimate (Initial Stock)
Stock as of 2012
Handguns
65,000,000
113,734,849
Ries
70,000,000
108,418,522
Shotguns
49,000,000
70,827,064
Misc.
8,000,000
8,972,118
All guns
192,000,000
301,952,553
Needles to say, these estimates are necessarily rough. Nonetheless, several facts are worth mentioning: The estimated stock is approximately 300 million rearms as of December 2012. Almost 50 millions additional guns has been consumed since 2008 in the United States (i.e: a 20-percent increase in the overall stock of rearms in circulation).
Figure S.2 presents the evolution of the
estimated stocks for the main category of guns from 2004 to 2012 and provides another interesting result: the consumption of handguns started to accelerate since 2007-2008 (it went up by 30 percent in ve years).
3
Increases in Applications for Permits to Carry Concealed Weapons and Gun Permits
For a group of selected states for which data were available, Table S.11 presents gures on applications for permits to carry concealed weapons and gun permits for the time period 2007-2009. The number of concealed weapon and rearm applications received by the state of Florida more than doubled in
13
two years.
Minnesota received 21,646 permit to carry applications in 2008, up 132 percent from
9,327 a year earlier. For some states this abrupt change was not only sizable in terms of the ow of permits and licensees but also in their stock: the 2009- 2008 ow in Arkansas's new carry concealed weapon licensees (i.e: 38,442 new licensees) represented 80 percent of the total stock of licensees as of December 2007. These gures provide an additional piece of evidence toward the existence of an Obama eect on rearms prevalence.
References [1] Charles, Kerwin K., and Jonathan Guryan, Prejudice and Wages: An Empirical Assessment of Becker's The Economics of Discrimination, Journal of Political Economy, vol. 116, no. 5, (2008).
[2] Cook, Phillip and Ludwig, J. (1997). Guns in America: Results of a Comprehensive Survey on Private Firearms Ownership and Use, abridged version published as a Research in Brief, National Institute of Justice, 1997, NCJ 165476
13 Florida
distributed 42,895 concealed weapon and rearm applications in November 2008 (an increase of 172
percent vs November 2007).
5
[3] Fisman, Raymond, Sheena S. Iyengar, Emir Kamenica, and Itamar Simonson, Racial Preferences in Dating, Review of Economic Studies Vol. 75 (2008):117132
[4] Kahneman, Daniel and Amos Tversky, Judgment under Uncertainty: Heuristics and Biases, Science, New Series, Vol. 185, No. 4157. (1974): 1124-1131
[5] Mas, Alexandre and Enrico Moretti, Racial Bias in the 2008 Presidential Election, American Economic Review: Papers & Proceedings 99 (2009): 323-329.
[6] Perry, Barbara. In The Name of Hate: Understanding Hate Crimes. New York: Routledge, 2001. [7] Rubenstein, William B. The Real Story of U.S. Hate Crime Statistics: An Empirical Analysis, Tulane Law Review, Vol. 78, (2004): 1213-1246.
[8] Saad, Lydia. State of the States: Consumer Condence. No economic oasis, little relief to be found across the country. Gallup on line article. January 29, 2009.
[9] United States Department of Justice. Federal Bureau of Investigation. UNIFORM CRIME REPORTING PROGRAM DATA [UNITED STATES]: OFFENSES KNOWN AND CLEARANCES BY ARREST, 2004. Compiled by the U.S. Dept. of Justice,
Federal Bureau of Investigation. ICPSR04459-v1. Ann Arbor,
MI: Inter-university Consortium for Political and Social Research [producer and distributor], 2006-08-18. http://doi.org/10.3886/ICPSR04459.v1 [10] United States Department of Justice. Federal Bureau of Investigation. UNIFORM CRIME REPORTING PROGRAM DATA [UNITED STATES]: OFFENSES KNOWN AND CLEARANCES BY ARREST, 2005. Compiled by the U.S. Dept. of Justice,
Federal Bureau of Investigation. ICPSR04721-v1. Ann Arbor,
MI: Inter-university Consortium for Political and Social Research [producer and distributor], 2007-06-08. http://doi.org/10.3886/ICPSR04721.v1 [11] United States Department of Justice. Federal Bureau of Investigation. UNIFORM CRIME REPORTING PROGRAM DATA [UNITED STATES]: OFFENSES KNOWN AND CLEARANCES BY ARREST, 2006. Compiled by the U.S. Dept. of Justice,
Federal Bureau of Investigation. ICPSR22400-v1. Ann Arbor,
MI: Inter-university Consortium for Political and Social Research [producer and distributor], 2008-05-20. http://doi.org/10.3886/ICPSR22400.v1 [12] United gram
States
Data
Arbor,
Department
[United
MI:
of
States]:
Inter-university
Justice. Oenses
Federal Known
Consortium
for
Bureau and
of
Investigation.
Clearances
Political
and
by
Social
Uniform
Arrest,
2007.
Research
Crime
Reporting
Pro-
ICPSR25101-v2.
Ann
[distributor],
2009-08-21.
http://doi.org/10.3886/ICPSR25101.v2 [13] United gram Arbor,
States
Data MI:
Department
[United
of
States]:
Inter-university
Justice. Oenses
Federal Known
Consortium
for
Bureau and
of
Investigation.
Clearances
Political
and
by
Social
Uniform
Arrest,
2008.
Research
Crime
Reporting
Pro-
ICPSR27648-v1.
Ann
[distributor],
2010-06-21.
http://doi.org/10.3886/ICPSR27648.v1 [14] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Oenses Known and Clearances by Arrest, 2009. ICPSR30766-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2011-09-30. http://doi.org/10.3886/ICPSR30766.v1 [15] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Oenses Known and Clearances by Arrest, 2010. ICPSR33526-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2012-06-19. http://doi.org/10.3886/ICPSR33526.v1 [16] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Oenses Known and Clearances by Arrest, 2011. ICPSR34586-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2013-04-26. http://doi.org/10.3886/ICPSR34586.v1 [17] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Oenses Known and Clearances by Arrest, 2012. ICPSR35021-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research[distributor], 2014-05-22. http://doi.org/10.3886/ICPSR35021.v1 [18] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data [United States]:
Supplementary Homicide Reports, 2004 . ICPSR04465-v1. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2006-08-18. http://doi.org/10.3886/ICPSR04465.v1
6
[19] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data [United States]:
Supplementary Homicide Reports, 2005. ICPSR04723-v1. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2007-06-11. http://doi.org/10.3886/ICPSR04723.v1 [20] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data [United States]:
Supplementary Homicide Reports, 2006. ICPSR22401-v1. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2008-07-08. http://doi.org/10.3886/ICPSR22401.v1 [21] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data [United States]:
Supplementary Homicide Reports, 2007. ICPSR25103-v1. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2009-06-10. http://doi.org/10.3886/ICPSR25103.v1 [22] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data [United States]:
Supplementary Homicide Reports, 2008. ICPSR27650-v2. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2010-05-10. http://doi.org/10.3886/ICPSR27650.v2 [23] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Supplementary Homicide Reports, 2009. ICPSR30767-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2011-08-04. http://doi.org/10.3886/ICPSR30767.v1 [24] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Supplementary Homicide Reports, 2010. ICPSR33527-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2012-06-19. http://doi.org/10.3886/ICPSR33527.v1 [25] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Supplementary Homicide Reports, 2011. ICPSR34588-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2013-04-26. http://doi.org/10.3886/ICPSR34588.v1 [26] United States Department of Justice. Federal Bureau of Investigation. Uniform Crime Reporting Program Data: Supplementary Homicide Reports, 2012. ICPSR35023-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research[distributor], 2014-04-16. http://doi.org/10.3886/ICPSR35023.v1
Figure 1: Change in AttitudesToward Gun Control
7
Table 1: State Characteristics State
Ln of #
Weakness
BCR/1k inhab
State Gun
Average Prejudice
Rep.
Gun
Leading
Income
Average
Large
Per Capita
Crime Rate
Obama
Mean
Std. Dev
Control
Margin
Ownership
Indicator
2007 ('000)
2007 2008
Eect
Alabama
1.59
0.30
0.85
0.76
0.15
0.52
-1.24
32.80
44.78
Y
Alaska
1.93
0.21
0.96
0.50
0.21
0.58
0.03
41.08
38.12
Y
Arizona
0.95
0.16
0.94
0.51
0.05
0.31
-2.29
34.37
48.17
N
Arkansas
1.68
0.27
0.94
0.70
0.00
0.55
-0.88
31.52
44.10
N
California
0.61
0.11
0.21
0.57
-0.14
0.21
-0.90
43.40
35.00
N
Colorado
1.52
0.22
0.84
0.39
0.00
0.35
-0.78
42.45
32.73
Y
Connecticut
1.31
0.26
0.46
0.68
-0.15
0.17
-0.57
55.63
27.06
Y
Delaware
0.52
0.26
0.78
0.65
-0.14
0.26
-1.73
39.93
41.74
Y
Florida
0.79
0.21
0.94
0.71
0.00
0.25
-2.33
39.04
48.21
Y
Georgia
1.03
0.29
0.93
0.69
0.07
0.40
-1.47
34.61
44.44
Y
Hawaii
-0.60
0.23
0.57
0.34
-0.22
0.09
-1.79
40.92
41.71
Y
Idaho
1.67
0.18
0.94
0.39
0.27
0.55
-2.58
32.84
24.08
N
Illinois
1.40
0.21
0.72
0.65
-0.16
0.20
-1.32
41.72
34.63
N
Indiana
1.05
0.31
0.92
0.63
0.09
0.39
-0.99
33.70
37.00
Y
Iowa
1.06
0.33
0.84
0.53
0.05
0.43
-0.23
35.75
28.07
N
Kansas
1.29
0.29
0.93
0.62
0.17
0.42
-0.75
37.41
39.60
N
Kentucky
3.60
0.14
0.98
0.82
0.09
0.48
-1.31
31.06
28.47
Y
Louisiana
1.43
0.33
0.98
0.86
0.05
0.44
0.20
35.34
46.42
Y
Maine
1.25
0.26
0.88
0.56
-0.12
0.41
-2.02
35.03
25.58
N
Maryland
0.24
0.23
0.47
0.64
-0.17
0.21
-1.60
46.92
41.10
Y
Massachusetts
0.44
0.25
0.46
0.52
-0.26
0.13
-0.03
49.64
28.36
N
Michigan
1.04
0.20
0.78
0.78
-0.09
0.38
-2.57
34.18
35.19
Y
Minnesota
1.45
0.21
0.89
0.46
-0.09
0.42
0.14
41.69
32.19
N
Mississippi
1.53
0.35
0.95
0.83
0.13
0.55
-0.08
29.54
33.59
N
Missouri
1.47
0.24
0.96
0.68
-0.01
0.42
-0.94
35.12
42.06
N
Montana
2.12
0.16
0.92
0.48
0.10
0.58
-0.98
33.93
29.57
N
Nebraska
0.92
0.32
0.90
0.55
0.23
0.39
-0.04
37.90
33.23
N
Nevada
1.04
0.27
0.89
0.67
-0.02
0.34
-3.84
40.93
43.50
Y
New Hampshire
1.47
0.16
0.89
0.54
-0.04
0.30
-0.49
42.67
21.39
N
New Jersey
-0.87
0.23
0.37
0.74
-0.12
0.12
-0.89
50.36
25.81
N
New Mexico
1.44
0.17
0.94
0.39
-0.06
0.35
-0.92
32.09
44.75
Y
New York
-0.05
0.20
0.49
0.71
-0.23
0.18
0.00
47.63
23.92
N
North Carolina
1.07
0.26
0.80
0.69
0.06
0.41
-1.17
34.87
45.33
Y
North Dakota
1.76
0.25
0.96
0.54
0.16
0.51
0.86
36.68
20.46
N
Ohio
0.92
0.25
0.87
0.78
-0.01
0.32
-1.49
35.17
37.79
N
Oklahoma
1.57
0.25
0.98
0.65
0.20
0.43
-0.54
34.30
39.98
Y
Oregon
1.34
0.20
0.82
0.47
-0.08
0.40
-1.02
35.74
36.77
Y
Pennsylvania
1.37
0.18
0.74
0.85
-0.07
0.35
-0.77
38.74
27.99
N
Rhode Island
-0.03
0.23
0.53
0.70
-0.26
0.13
-1.78
40.15
29.70
Y
South Carolina
1.19
0.28
0.91
0.76
0.12
0.42
-1.54
31.92
50.12
Y
South Dakota
1.80
0.31
0.94
0.50
0.11
0.57
0.15
36.43
18.34
N
Tennessee
1.37
0.35
0.93
0.73
0.05
0.44
-0.80
34.16
48.03
Y
Texas
1.12
0.24
0.91
0.62
0.13
0.36
-0.06
36.84
45.63
Y
Utah
1.69
0.41
0.96
0.30
0.31
0.44
-0.69
31.80
36.57
Y
Vermont
1.05
0.22
0.91
0.59
-0.21
0.42
-0.55
37.73
25.61
N
Virginia
1.05
0.25
0.82
0.59
0.03
0.35
-0.78
43.16
27.55
N
Washington
1.36
0.17
0.82
0.50
-0.11
0.33
-0.76
41.92
42.27
Y
West Virginia
1.87
0.26
0.96
1.00
0.01
0.55
-0.89
30.12
28.21
N
Wisconsin
0.99
0.29
0.88
0.63
-0.06
0.44
-0.91
36.99
30.80
N
Wyoming
2.00
0.19
0.91
0.48
0.26
0.60
-0.10
46.73
30.27
Y
BCR: Background check reports. See main text for variable denitions -
8
Table 2: Obama Eect and Gun-Control Fear (Discrete Interaction Terms) Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants Panel A: F(Obama) = Obama
F (Obama) F (Obama) * Low State Gun Control State FE Season FE Year FE Month FE Add. Interaction Terms
Panel B: F(Obama) = P (Obama)
(1) 0.213*** (0.0237)
(2) 0.206*** (0.0273)
(3)
(1) 0.482*** (0.0456)
(2) 0.473*** (0.0495)
(3)
0.115**
0.0746*
0.0671*
0.165**
0.110**
0.100**
(0.0555) Y Y Y N N
(0.0385) Y Y Y N Y
(0.0363) Y N N Y Y
(0.0775) Y Y Y N N
(0.0512) Y Y Y N Y
(0.0496) Y N N Y Y
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008. All specications included constant and the leading indicator as control (not reported). Sample size is 1,700 state-month observations (50 states). The leading indicator for each state predicts the six-month growth rate of the coincident index. Columns (1) and (2) in panel B include the Probability of other democract winning the election as control. The additional interaction terms are F (Obama) * High Republican Margin and F (Obama) * Low Gun Ownership. Low and High are dened with respect to median values. -
Table 3: Obama Eect and Race Bias (Discrete Interaction Terms) Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants Panel A: F(Obama) = Obama
F (Obama) F (Obama) * High Prejudice State FE Season FE Year FE Month FE Add. Interaction Terms
Panel B: F(Obama) = P (Obama)
(1) 0.165*** (0.0404)
(2) 0.105*** (0.0361)
(3)
(1) 0.416*** (0.0675)
(2) 0.339*** (0.0640)
(3)
0.149**
0.177***
0.167***
0.202**
0.236***
0.225***
(0.0593) Y Y Y N N
(0.0460) Y Y Y N Y
(0.0498) Y N N Y Y
(0.0866) Y Y Y N N
(0.0705) Y Y Y N Y
(0.0735) Y N N Y Y
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008. All specications included constant and the leading indicator as control (not reported). Sample size is 1,700 state-month observations (50 states). The leading indicator for each state predicts the six-month growth rate of the coincident index. Columns (1) and (2) in panel B include the Probability of other democract winning the election as control. The additional interaction terms are F (Obama) * High Republican Margin and F (Obama) * Low Gun Ownership. Low and High are dened with respect to median values. -
9
-
Table 4: Obama Eect, Gun Control Fear, and Race Bias (Discrete Interaction Terms) Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants Panel A: F(Obama) = Obama
F (Obama) F (Obama) * Low State Gun Control F (Obama) * High Prejudice State FE Season FE Year FE Month FE Add. Interaction Terms
Panel B: F(Obama) = P (Obama)
(1) 0.120*** (0.0384)
(2) 0.0987*** (0.0365)
(3)
(1) 0.353*** (0.0665)
(2) 0.329*** (0.0636)
(3)
0.114**
0.0543***
0.0504**
0.164**
0.0828***
0.0771**
(0.0461)
(0.0204)
(0.0233)
(0.0642)
(0.0265)
(0.0308)
0.149***
0.173***
0.164***
0.202***
0.230***
0.220***
(0.0459) Y Y Y N N
(0.0439) Y Y Y N Y
(0.0481) Y N N Y Y
(0.0680) Y Y Y N N
(0.0675) Y Y Y N Y
(0.0708) Y N N Y Y
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008. All specications included constant and the leading indicator as control (not reported). Sample size is 1,700 state-month observations (50 states). The leading indicator for each state predicts the six-month growth rate of the coincident index. Columns (1) and (2) in panel B include the Probability of other democract winning the election as control. The additional interaction terms are F (Obama) * High Republican Margin and F (Obama) * Low Gun Ownership. Low and High are dened with respect to median values.
Table 5: Omitting Election Aftermath Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants
P (Obama) P (Obama) * Weakness State Gun Control
(1) 0.576*** (0.0901)
P (Obama) * Prejudice State FE Season FE Year FE Month FE Additional Interaction Terms
Y Y Y N N
(2) 0.0518 (0.264) 0.705*** (0.175) 0.324 (0.372) Y Y Y N Y
(3) 0.705*** (0.170) 0.320 (0.388) Y N N Y Y
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008. All specications included constant and the leading indicator as control (not reported). Sample size is 1,050 state-month observations (50 states). The leading indicator for each state predicts the six-month growth rate of the coincident index. Columns (1) and (2) include the Probability of other democract winning the election as control. The additional interaction terms are P (Obama) * Average Republican Margin and P (Obama) * Gun Ownership. -
10
Table 6: Using Polls Data Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants
P (Obama) P (Obama) * Weakness State Gun Control
(1) 0.487*** (0.0397)
P (Obama) * Average Prejudice State FE Season FE Year FE Month FE Add. Interaction Terms Period
Y Y Y N N
(2) 0.130 (0.170)
(3)
0.444***
(5) -0.164 (0.184)
(6)
0.436***
0.704***
0.705***
(0.133)
(0.137)
(0.133)
(0.114)
0.449**
0.419*
0.451
0.452
(0.224) (0.234) Y Y Y N Y N N Y Y Y Mar-2007 / Dec-2009
(4) 0.509
Y Y Y N N
(0.332) (0.334) Y Y Y N Y N N Y Y Y Mar-2007 / Nov-2008
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008 and include the leading indicator as control (not reported). The leading indicator for each state predicts the six-month growth rate of the coincident index. Sample size in rst (last) 3 columns is 1700 (1050) state-month observations (with 50 states). Columns (1), (2), (4), and (5) include the probability of other democrat winning the election. The additional interaction terms are P (Obama) * Average Republican Margin and P (Obama) * Gun Ownership. -
Table 7: Using Alternative Measures for Prejudice Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants
P (Obama) * Alternative Meassure for Prejudice P (Obama) * Weakness State Gun Control Prejudice Measure State FE Month FE Additional Interaction Terms Observations
(1)
(2)
(3)
(4)
20.59
2.664***
0.789**
0.236**
(31.45)
(0.669)
(0.339)
(0.120)
0.526***
0.489***
0.482***
0.411***
(0.145) Hate Crimes in 2008 (FBI) Y Y Y 1,700
(0.114) Fisman et al (2008) Y Y Y 1,428
(0.126) Mas and Moretti (2008) Y Y Y 1,428
(0.142) Charles and Guryan (2008) Y Y Y 1,462
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008 and include the leading indicator as control (not reported). The leading indicator for each state predicts the six-month growth rate of the coincident index. Hates Crimes account for the number of racially motivated hate crimes per million of people covered by agencies reporting to the FBI. The additional interaction terms are P (Obama) * Average Republican Margin and P (Obama) * Gun Ownership.
11
Table 8: Further Robustness Check Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants
P (Obama) * Weakness State Gun Control
P (Obama) * Average Prejudice
(1)
(2)
(3)
(4)
(5)
(6)
0.384***
0.608***
0.477***
0.427***
0.460***
0.495***
(0.144)
(0.174)
(0.137)
(0.122)
(0.135)
(0.0936)
0.882***
0.510
0.403
0.383
0.388
0.374
(0.241)
(0.344)
(0.285)
(0.257)
(0.250)
(0.272)
P (Obama) * Economic
-0,065
Condence Index 2008
(0.057) P (Obama) * Google's Volume Search for
0.278
Stimulus and Economic Stimulus (0.213) P (Obama) * Relative
9,73
Black-White Population
(2.13) P (Obama) * Log of
0.409
Income Per Capita
(0.384) P (Obama) * Crime
0,107
Rate (Average 2006-2007)
(0.487) P (Obama) * Share of
0,294
Rural Population
(0.455) Robustness
Includes Interaction
Omits
Includes Interaction
Includes Interaction
Includes Six
Includes Interactions
Terms for Prob
Southern
Economic
Google's Volume
Lags of Log of
for Income, Crime,
Other Democrat
States
Condence Index
Search
Monthly Crime
Race, and Rural
State FE
Y
Y
Y
Y
Y
Y
Month FE
Y
Y
Y
Y
Y
Y
Additional Interaction Terms
Y
Y
Y
Y
Y
Y
1.700
1,156
1.700
1.700
1,632
1.700
50
34
50
50
48
50
State-Month Observations Number of States
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008 and include the leading indicator as control (not reported). The leading indicator for each state predicts the six-month growth rate of the coincident index. The additional interaction terms are P (Obama) * Average Republican Margin and P (Obama) * Gun Ownership. Smaple size in column (5) drops due to the lack of monthly crime data for Alabama and Florida Economic Condence Index 2008 comes from Gallup and it is based on 2 questions regarding public perception of the current and future economic conditions. Income Per Capita is average quarterly of the current and future economic conditions. personal income in 2007. Relative Black-White Population is the ratio black to white population in 2008. Share of Rural Population is based on 2000's population. -
12
Table 9: Alternative Dependent Variable Dependent variable: Log of Criminal Firearm
Log of Google
Background Checks
Searches for Gun
(1) 0.468*** (0.138) 0.432 (0.266) Y Y Y Y 1,462 43
(2) 0.135*** (0.0449) 0.158* (0.0875) Y Y Y Y 1,462 43
P (Obama) * Weakness State Gun Control P (Obama) * Average Prejudice State FE Month FE Additional Interaction Terms Weighted by Population State-Month Observations Number of States
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008 and include the leading indicator as control (not reported). The leading indicator for each state predicts the six-month growth rate of the coincident index. The additional interaction terms are P (Obama) * Average Republican Margin and P (Obama) * Gun Ownership. The following states are not included in the regressions due to lack of google search data: Delaware, Montana, North Dakota, South Dakota, Vermont, West Virginia, and Wyoming.
Table 10: Regional Distribution of Obama Eect Dependent variable: Log of Criminal Firearm Background Checks per 1,000 inhabitants
(1) 0.0573 (0.180) 0.273*** (0.0580) 0.301*** (0.0467) 0.258*** (0.0526) 0.329*** (0.0382) 0.383*** (0.0868) 0.298*** (0.0421) 0.272*** (0.0501) 0.0605* (0.0315)
F (Obama) * New England F (Obama) * Mid-Atlantic F (Obama) * East North Central F (Obama) * West North Central F (Obama) * South Atlantic F (Obama) * East South Central F (Obama) * West South Central F (Obama) * Mountain F (Obama) * Pacic
(2) 0.149 (0.278) 0.442*** (0.0926) 0.491*** (0.0726) 0.443*** (0.0795) 0.533*** (0.0763) 0.596*** (0.126) 0.498*** (0.0868) 0.485*** (0.0998) 0.161** (0.0751)
*** signicant at the 1 percent.** signicant at the 5 percent.* signicant at 10 percent. Standard errors adjusted for two-way clustering within states and within months in parentheses. All models are weighted by state population in 2008. Sample size is 1700 state-month observations (with50 states). All specications include State FE, Season FE, and Year FE. I also control for the leading indicator of the State Coincident Index. F (Obama) is Obama Dummy in Column (1), and Prob (Obama) in column (2). -
13
Table 11: CCW and Gun Permit Statistics State
Figure
2007
2008
13,945 70% 17,695 89% 10,042 42% 86,269 14% 121,219 79% 25,996 19% 5,772 -31% 6,617 5% 33,411 49% 21,646 132% 13,325 9% 3,857 18% 33,864 53% 3,373 68% 14,631 41% 62,185 27% 85,973 -2% 37,443 29% 55,864 35%
Arkansas
CCW New Licenses
8,214
Colorado
CCW Permits Issued by Colorado Sheris
9,370
New Gun Permit Issued
7,091
Florida
CCW Applications (New and Renewal)
75,679
Georgia
Firearm Licenses Applications
67,640
Hawaii
Total Number of Firearms Registered
21,784
Kansas
CCW Licenses Issued
8,410
Louisiana
CCW Permits Issued
6,320
Michigan*
Concealed Pistol Licenses Applications
22,403
Minessota
Permit to Carry Applications
9,327
Pistol Licenses Issued
12,232
CCW Permits Issued (New and Renewal)
3,263
Concealed Handgun Licenses Issued
22,103
Handgun Safety Tests
2,004
CCW Permits Issued
10,379
Handgun Carry Permit Issued
48,991
Texas
Concealed Handgun Licenses Issued
87,391
Utah
CCW Permits Issued
29,134
Concealed Handgun Permits Issued
41,472
Connecticut
New York** North Dakota Ohio Rhode Island South Carolina Tennessee
Virginia
2009
24,497 76% 27,000 53% 16,283 62% 166,964 94% 92,452 -24% 33,678 30% 7,545 31% 9,210 39% 73,105 119% 22,378 3% 18,577 39% 5,357 39% 56,691 67% . . 34,879 138% 116,607 88% 138,768 61% 76,324 104% 70,434 26%
Notes: CCW stands for Carrying a Concealed Weapon. * Figures are for period July-June. ** Excluding New York City Percentages below each annual gure represents its growth rate with respect to the previous year
14
Figure 2: Evolution of Stock of Guns (1994-2012)
15