DataPDF Available
Distributive Politics and Crime
Online Appendix
Masataka Harada
Daniel M. Smith
August 3, 2021
Abstract
We examine whether and how intergovernmental scal transfers reduce crime, an im-
portant but understudied aspect of distributive politics. Estimating the causal eect
of redistribution on crime is complicated by the problem of simultaneity: transfers
may be targeted precisely where crime is a problem. Our research design takes advan-
tage of municipality-level panel data from Japan spanning a major electoral system
reform that reduced the level of malapportionment across districts. This provides
an opportunity to use the change in malapportionment as an instrumental variable,
as malapportionment aects redistribution outcomes, but the change caused by the
reform is orthogonal to local crime rates. Naïve OLS estimates show negligible (near
zero) eects of transfers on crime, whereas the IV results reveal larger negative ef-
fects. This nding supports the argument that redistribution can reduce crime, and
introduces a new perspective on the relationship between Japan’s well-known pattern
of distributive politics and its comparatively low crime rates.
Keywords: distributive politics, crime, malapportionment, instrumental variable,
Japan
This online appendix contains supplementary information and analyses referenced in the main text of
the article appearing in the Journal of Political Institutions and Political Economy.
Department of Economics, Fukuoka University. 8-19-1 Nanakuma, Jonan-ku, Fukuoka 814-0180,
Japan. Email: masatakaharada@gmail.com. Corresponding author.
Department of Political Science and School of International and Public Aairs, Columbia University.
420 W. 118th Street, 915 International Aairs Building, New York, NY 10027, United States. Email:
dms2323@columbia.edu.
A Supplementary Tables and Figures
Table A.1: Total reported penal code oenses in Japan, 1993-1999
1993 1994 1995 1996 1997 1998 1999
Homicide 1,233 1,279 1,281 1,218 1,282 1,388 1,265
Robbery 2,466 2,684 2,277 2,463 2,809 3,426 4,237
Injury 18,306 18,097 17,482 17,876 19,288 19,476 20,233
Assault 6,576 6,112 6,190 6,469 7,254 7,367 7,792
Intimidation 940 1,019 943 904 1,040 971 995
Fraud 47,341 52,047 45,923 49,394 49,426 48,279 43,431
Extortion 11,225 11,266 11,207 12,226 12,947 13,900 14,768
Embezzlement (a) 1,679 1,875 1,632 1,621 1,569 1,355 1,229
Embezzlement (b) 59,820 66,629 59,512 58,592 58,955 64,025 67,635
Rape 1,611 1,616 1,500 1,483 1,657 1,873 1,857
Forcible indecency 3,581 3,580 3,644 4,025 4,398 4,251 5,346
Arson 1,754 1,741 1,710 1,846 1,936 1,566 1,728
Obstruction of duty 965 1,113 1,188 1,268 1,434 1,395 1,531
Burglary 11,942 11,213 11,009 11,246 12,281 13,308 14,549
Damage to property 30,707 30,119 31,231 36,406 41,064 46,009 53,552
Total reported crimes 200,146 210,390 196,729 207,037 217,340 228,589 240,148
Notes: Data are from the National Police Agency of Japan. Embezzlement (a) excludes embez-
zlement of lost property; (b) is for embezzlement of lost property. Obstruction of duty is for the
obstruction of the performance of duty by a public ocial (e.g., a police ocer). Burglary refers
to breaking into a residence.
1
Table A.2: Descriptive statistics of the data sample
Variable N Mean SD Min. Max.
Crimes per 1,000 Residents (log) 1,376 2.59 .413 1.22 5.09
Total Unemployment Rate (log) 1,364 -3.19 .283 -4.10 -1.88
Male Unemployment Rate (log) 1,364 -3.09 .284 -3.97 -1.56
Female Unemployment Rate (log) 1,364 -3.34 .305 -4.32 -2.15
Taxable Income Per Capita (log) 1,376 0.348 .238 -.372 1.33
Local allocation tax per capita (log) 1,376 -3.33 1.59 -9.35 -.695
Malapportionment (log) 1,376 1.16 .371 .551 1.94
Population (log) 1,376 11.3 .913 8.80 15.0
Ratio of Population Aged 15 and Younger 1,376 .159 .0189 .0877 .240
Ratio of Population Aged 65 and Older 1,376 0.163 .0429 .0629 .295
Population Density (log) 1,376 6.75 1.37 3.10 9.84
Notes: Only the year 1996 and 1997 are used to calculate the descriptive statistics. The number
of observations vary slightly depending on the years used and variables included in the models.
Local allocation tax data are from Horiuchi and Saito (2003), who use socioeconomic variables
from the 1995 census; for subsequent years, we collected corresponding data from the 2000 census,
using interpolation to ll in missing years. Crime data are from annually reported ocial crime
statistics, Hanzai Tōkei. When a single police district contains multiple municipalities, we use the
population-weighted crime statistic as an approximation. However, if a municipality is covered
by multiple police districts, we exclude all aected municipalities. This process drops eight cities
in Tokyo (but none of Tokyo’s 23 wards). Other socioeconomic variables are collected from
Ocial Statistics of Japan (http://www.e-stat.go.jp/). Electoral variables are from the Reed-
Smith Japanese House of Representatives Elections Dataset (Reed and Smith, 2018).
2
Table A.3: Complete rst-stage results: regression of per capita local allocation tax on malappor-
tionment
DV: Local allocation tax per capita (log)
(1) (2)
Malapportionment (log) .248 .206
(.0533) (.0530)
Year 1997 .224 .0409
(.0329) (.0792)
Population (log) .825
(5.42)
Ratio of population aged 15 and younger .2.12
(8.86)
Ratio of population aged 65 and older 27.1
(9.20)
Population density (log) 5.43
(4.98)
Municipality xed eects ✓ ✓
Within R2.157 .172
Cragg-Donald Wald F statistic 41.5 26.3
Kleibergen-Paap rk Wald F statistic 21.7 15.1
Number of units (municipalities) 688 688
Number of observations 1,376 1,376
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Within
R2estimated separately with xtreg command. Standard errors in parentheses are clustered by
single-member district (SMD) and year.
3
-4 -2 0 2 4 6
1st Diff. Population (log)
-8 -6 -4 -2 0 2
1st Diff. Pop. < 15
-4 -2 0 2 4 6
1st Diff. Pop. > 65
-4 -2 0 2 4 6
1st Diff. Pop. Density(log)
Figure A.1: Kernel density plots for the socioeconomic covariates included in the IV regression
Notes: Balances are compared between the cities in which the change in the malapportionment is
in the upper 75th percentile (red solid line), the 50–75th percentile (orange long-dashed line), the
25–50th percentile (yellow dashed line), and those in the lower 25th percentile (green short-dashed
line). All covariates in the gure are transformed by taking the rst dierence between the year
1996 and 1997.
4
Table A.4: First and second-stage results using the data from 1995 and 1996 to check trend eects
Stage: 1st Stage 2nd Stage
DV: Local allocation tax Crimes per 1,000
per capita (log) residents (log)
(1) (2)
Local Allocation Tax Per Capita (log) -2.12
(6.58)
Malapportionment (log) -.478
(2.02)
Year 1996 -.00071 .0310
(.0449) (.0688)
Population (log) -5.06 -8.81
(7.80) (34.0)
Ratio of Population Aged 15 and Younger 16.8 37.4
(5.70) (112)
Ratio of Population Aged 65 and Older 19.8 40.5
(7.08) (132)
Population Density (log) 9.07 17.5
(7.58) (60.5)
Municipality xed eects ✓ ✓
Cragg-Donald Wald F statistic 0.055
Kleibergen-Paap rk Wald F statistic 0.056
AR 95% Condence Set [−∞,]
Number of units (municipalities) 688
Number of observations 1,376
Notes: This analysis uses variables measured in 1995 and 1996 rather than 1996 and 1997 (as in
the main analysis). Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010).
Standard errors in parentheses are clustered by SMD and year. The AR α%condence set is calcu-
lated with Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based
on Anderson and Rubin (1949), where the condence sets are estimated with Wald/Minimum
Distance tests with a grid search of 2,000 times.
5
Table A.5: Complete second-stage results: regression of logged crime rates on per capita local
allocation tax using malapportionment as an IV (with comparison to naïve OLS)
DV: Crimes per 1,000 residents (log)
OLS IV
(1) (2) (3) (4)
Local allocation tax per capita (log) -.0351 -.0347 -.220 -.249
(.0122) (.0119) (.103) (.122)
Year 1997 .0529 -.0609 .0704 -.0858
(.00639) (.0387) (.0139) (.0451)
Population (log) 1.44 1.78
(3.38) (3.91)
Ratio of population aged 15 and younger -2.61 -1.50
(3.41) (3.59)
Ratio of population aged 65 and older 18.1 25.9
(6.21) (8.52)
Population density (log) -1.25 .330
(3.32) (3.97)
Municipality xed eects ✓ ✓
AR 95% Condence Set [-.453,-.030] [-.541,-.026]
Number of units (municipalities) 688 688 688 688
Number of observations 1,376 1,376 1,376 1,376
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard
errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated with
Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests
with a grid search of 2,000 times.
6
Table A.6: First and second-stage results using values from prior years (1995 and 1996) for
dependent variable as a placebo test
Stage: 1st Stage 2nd Stage
DV: Local allocation tax Crimes per 1,000
per capita (log) residents (log)
(1) (2)
Local Allocation Tax Per Capita (log) .061
(.090)
Malapportionment (log) .208
(.053)
Year 1996 .036 .037
(.082) (.031)
Population (log) 1.01 2.38
(5.43) (1.33)
Ratio of Population Aged 15 and Younger .981 -3.37
(9.93) (3.34)
Ratio of Population Aged 65 and Older 27.5 -6.35
(9.22) (5.27)
Population Density (log) 5.42 -2.57
(4.98) (1.33)
Municipality xed eects ✓ ✓
Cragg-Donald Wald F statistic 26.34
Kleibergen-Paap rk Wald F statistic 15.15
AR 95% Condence Set [-.116,.261]
Number of units (municipalities) 684
Number of observations 1,368
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard
errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated with
Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests
with a grid search of 2,000 times.
7
Table A.7: First and second-stage results using extended set of control variables
Stage: 1st Stage 2nd Stage
DV: Local allocation tax Crimes per 1,000
per capita (log) residents (log)
(1) (2)
Local Allocation Tax Per Capita (log) -.325
(.175)
Malapportionment (log) 0.152
(.0463)
Year 1996 .00916 .122
(.0834) (.0572)
Population (log) -7.95 -1.42
(11.1) (5.41)
Ratio of Population Aged 15 and Younger -8.55 -3.12
(8.35) (4.88)
Ratio of Population Aged 65 and Older 11.4 21.7
(8.81) (7.78)
Population Density (log) 10.5 2.68
(10.6) (5.62)
Ratio of Workers in Primary Sector 3.57 -1.55
(4.11) (3.13)
Ratio of Workers in Tertiary Sector 8.19 2.54
(3.74) (2.66)
Population Density (DID) -1.53 -.601
(1.17) (.884)
Municipality Fiscal Strength Index -4.50 -1.32
(.890) (.814)
District Magnitude -.00322 -.0132
(.00903) (.00767)
Total Number of Wins for -.0287 -.00931
Govt. Coal. Candidates (log) (.0138) (.00954)
Cabinet Experiences for .0141 .0164
Govt. Coal. Candidates (.0129) (.00853)
Municipality xed eects ✓ ✓
Cragg-Donald Wald F statistic 14.1
Kleibergen-Paap rk Wald F statistic 10.8
AR 95% Condence Set [-.778,-.011]
Number of units (municipalities) 686
Number of observations 1,372
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard
errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated with
Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests
with a grid search of 2,000 times.
8
Table A.8: First-stage results: regression of per capita local allocation tax on malapportionment
and battleground district as two IVs
Dependent Variable : Local allocation tax per capita (log)
Vote Share Margin for Battleground 0.5% 1% 2%
(1) (2) (3)
Malapportionment (log) .207 .202 .208
(.0532) (.0533) (.0529)
Dummy for Battleground District .123 .0831 .0622
(.0599) (.0442) (.0285)
Year 1997 .0714 .0589 .0524
(.0813) (.0795) (.0772)
Population (log) .438 1.35 1.28
(5.08) (5.10) (5.38)
Ratio of Population Aged 15 and Younger 2.61 2.36 4.11
(8.83) (8.83) (8.80)
Ratio of Population Aged 65 and Older 22.6 23.8 25.9
(9.34) 9.35 9.04
Population Density (log) 5.14 4.60 4.88
(4.62) (4.63) (4.91)
Municipality xed eects ✓ ✓
Cragg-Donald Wald F statistic 16.8 16.6 16.6
Kleibergen-Paap rk Wald F statistic 9.44 9.68 11.2
Number of units (municipalities) 686 686 686
Number of observations 1,372 1,372 1,372
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010) with CUE option.
Standard errors in parentheses are clustered by SMD and year. The dummy for battleground district is
coded as 1 if the seat-adjusted dierence in vote share (vote share dierence ×seat) between a marginal
candidate of the governing party coalition and an opposition party candidate is less than 0.5%, 1%, or 2%,
respectively.
9
Table A.9: Second-stage results: regression of logged crime rates on per capita local allocation
tax using malapportionment and battleground district as two IVs
Dependent Variable : Crimes per 1,000 Residents (log)
Vote Share Margin for Battleground 0.5% 1% 2%
(4) (5) (6)
Local allocation tax per capita (log) -.228 -.326 -.226
(.099) (.115) (.108)
Year 1997 -.0838 -.0925 -.0845
(.0434) (.0451) (.0448)
Population (log) 1.86 1.34 1.95
(3.86) (4.05) (3.85)
Ratio of Population Aged 15 and Younger -1.55 -2.35 -1.37
(3.54) (3.77) (3.47)
Ratio of Population Aged 65 and Older 25.1 28.3 25.1
(7.96) (8.32) (8.32)
Population Density (log) .133 1.09 .0817
(3.88) (4.08) (3.89)
Municipality xed eects ✓ ✓
P-value for Hansen J statistic .827 .316 .753
AR 95% Condence Set [-.503, .011] [-.634,-.118] [-.535, .023]
AR 90% Condence Set [-.461,-.020] [-.559,-.160] [-.487,-.009]
Number of units (municipalities) 686 686 686
Number of observations 1,372 1,372 1,372
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010) with CUE option.
Standard errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated
with Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests with a grid
search of 2,000 times. The dummy for battleground district is coded as 1 if the seat-adjusted dierence
in vote share (vote share dierence ×seat) between a marginal candidate of the governing party coalition
and an opposition party candidate is less than 0.5%, 1%, or 2%, respectively.
10
Table A.10: Second-stage results of the regression of logged crime rates on per capita local alloca-
tion tax: (1) original, (2) excluding cities where the headquarters of designated crime syndicates
are located, and (3) excluding cities that held local elections between FY 1996-97
Dependent Variable: Crimes per 1,000 Residents (log)
Type of Robustness Check Original Yakuza HQ Local Elec.
(1) (2) (3)
Local allocation tax per capita (log) -.249 -.286 -.259
(.122) (.133) (.124)
Year 1997 -.0858 .088 -.0859
(.0451) (.046) (.0456)
Population (log) 1.78 2.35 1.85
(3.91) (3.96) (3.91)
Ratio of Population Aged 15 and Younger -1.50 -.545 -1.79
(3.59) (3.68) (3.64)
Ratio of Population Aged 65 and Older 25.9 27.6 25.9
(8.52) (9.04) (8.61)
Population Density (log) .330 .118 .328
(3.97) (3.97) (3.97)
Municipality xed eects ✓ ✓
Cragg-Donald Wald F statistic 41.5 23.3 25.5
Kleibergen-Paap rk Wald F statistic 21.7 14.6 14.7
AR 95% Condence Set [-.541, -.026] [-.609,-.044] [-.559, -.034]
Number of units (municipalities) 688 671 679
Number of observations 1,376 1,342 1,358
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard errors in
parentheses are clustered by SMD and year. The AR α%condence set is calculated with Stata’s ado
program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson and Rubin (1949),
where the condence sets are estimated with Wald/Minimum Distance tests with a grid search of 2,000
times. We select the headquarters of crime syndicates that were designated by Anti-Organized Crime Law
before 1996 and still exist as of June 14, 2021 (Iwate Prefectural Council for Eliminating Gangsters, 2021).
Excluded cities where the headquarters of a designated crime syndicate (yakuza) was located are Kobe,
Minato-ku (Tokyo), Kitakyushu, Naha, Kyoto, Hiroshima, Shimonoseki, Kagoshima, Kasaoka, Kurume,
Takamatsu, Ichihara, Onomichi, Tagawa, Toshima-ku (Tokyo), Osaka, and Taito-ku (Tokyo). Excluded
cities holding local elections between the scal years of 1996 and 1997 are Itoman, Kushiro, Onojo, Hikone,
Kamifukuoka, Komae, Otsu, Nakatsugawa, Komoro.
11
Table A.11: First and second-stage results: regression of logged crime rates on per capita local
allocation tax including towns and villages
Stage: 1st Stage 2nd Stage
DV: Local allocation tax Crimes per 1,000
per capita (log) residents (log)
(1) (2)
Local Allocation Tax Per Capita (log) -.507
(.346)
Malapportionment (log) .0742
(.0182)
Year 1996 -.0668 -.0626
(.0164) (.0255)
Population (log) .796 -2.09
(.383) (1.13)
Ratio of Population Aged 15 and Younger 2.09 3.32
(1.55) (3.20)
Ratio of Population Aged 65 and Older 4.05 6.59
(1.38) (3.07)
Population Density (log) .0860 1.51
(.173) (.757)
Municipality xed eects ✓ ✓
Cragg-Donald Wald F statistic 42.6
Kleibergen-Paap rk Wald F statistic 16.7
AR 95% Condence Set [-1.29,.160]
Number of units (municipalities) 3,242
Number of observations 6,484
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard
errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated with
Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests
with a grid search of 2,000 times.
12
Table A.12: Complete second-stage results: regression of logged total unemployment rates and
logged male unemployment rates on logged per capita local allocation tax using malapportionment
as an IV (with comparison to OLS)
DV: Total Unemp. Rate (log) Male Unemp. Rate (log)
Estimation Method: OLS IV OLS IV
(1) (2) (3) (4)
Local allocation tax per capita (log) -.00708 -.0670 -.00452 -.0489
(.00191) (.0241) (.00196) (.0234)
Year 1997 .0341 .0268 .0325 .0270
(.00607) (.00776) (.00669) (.00794)
Population (log) -1.30 -1.19 -1.18 -1.10
(.430) (.569) (.430) (.505)
Ratio of Population Aged 15 and Younger -4.42 -4.20 -4.87 -4.71
(.681) (.827) (.717) (.766)
Ratio of Population Aged 65 and Older -3.51 -.1.31 -3.64 -2.01
(.891) (1.31) (.983) (1.37)
Population Density (log) .949 1.39 .985 1.31
(.415) (.570) (.411) (.515)
Municipality xed eects ✓ ✓ ✓ ✓
Cragg-Donald Wald F statistic n/a 26.4 n/a 26.4
Kleibergen-Paap rk Wald F statistic n/a 15.2 n/a 15.2
AR 95% Condence Set n/a [-.129,-.026] n/a [-.107,-.008]
Number of units (municipalities) 682 682 682 682
Number of observations 1,364 1,364 1,364 1,364
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard
errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated with
Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests
with a grid search of 2,000 times. Unemployment rates are interpolated from the 1995 and 2000
censuses.
13
Table A.13: Complete second-stage results: regression of logged female unemployment rates and
logged per capita taxable income on logged per capita local allocation tax using malapportionment
as an IV (with comparison to OLS)
DV: Female Unemp. Rate (log) Taxable Income P.C. (log)
Estimation Method: OLS IV OLS IV
(5) (6) (7) (8)
Local allocation tax per capita (log) -.0124 -.107 -.0127 -.0126
(.00259) (.0290) (.00226) (.0100)
Year 1997 .0347 .0232 .0306 .0306
(.00580) (.00855) (.00424) (.00408)
Population (log) -1.66 -1.49 2.90 2.90
(.505) (.785) (2.10) (2.09)
Ratio of Population Aged 15 and Younger -3.82 -3.47 -.0607 -.0611
(.758) (1.11) (.449) (.448)
Ratio of Population Aged 65 and Older -2.65 .807 -.258 -.260
(.825) (1.33) (.781) (.763)
Population Density (log) .961 1.66 -3.10 -3.10
(.490) (.760) (2.05) (2.06)
Municipality xed eects ✓ ✓ ✓ ✓
Cragg-Donald Wald F statistic n/a 26.4 n/a 26.3
Kleibergen-Paap rk Wald F statistic n/a 15.2 n/a 15.1
AR 95% Condence Set n/a [-.184,-.059] n/a [-.033,.008]
Number of units (municipalities) 682 682 688 688
Number of observations 1,364 1,364 1,376 1,376
Notes: Estimates are obtained using Stata’s ado program xtivreg2 (Schaer, 2010). Standard
errors in parentheses are clustered by SMD and year. The AR α%condence set is calculated with
Stata’s ado program weakiv (Finlay, Magnusson and Schaer, 2013), originally based on Anderson
and Rubin (1949), where the condence sets are estimated with Wald/Minimum Distance tests
with a grid search of 2,000 times. Unemployment rates are interpolated from the 1995 and 2000
censuses.
14
References
Anderson, T. W. and Herman Rubin. 1949. “Estimation of the Parameters of a Single
Equation in a Complete System of Stochastic Equations.The Annals of Mathematical
Statistics 20(1):46–63.
Finlay, Keith, Leandro Magnusson and Mark E. Schaer. 2013. “WEAKIV:
Stata module to perform weak-instrument-robust tests and condence inter-
vals for instrumental-variable (IV) estimation of linear, probit and tobit mod-
els.” Statistical Software Components, Boston College Department of Economics.
https://ideas.repec.org/c/boc/bocode/s457684.html.
Horiuchi, Yusaku and Jun Saito. 2003. “Reapportionment and Redistribution: Conse-
quences of Electoral Reform in Japan.American Journal of Political Science 47(4):669–
682.
Iwate Prefectural Council for Eliminating Gangsters. 2021. “Status of Designated Crime
Syndicates (Shitei Bouryokudan no Shitei Jyōkyō).http://www.rnac.ne.jp/~boutui/
map.html. Accessed: 2021-06-14.
Reed, Steven R. and Daniel M. Smith. 2018. “The Reed-Smith Japanese
House of Representatives Elections Dataset.” Harvard Dataverse, V1,
https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/QFEPXD.
Schaer, Mark E. 2010. “xtivreg2: Stata module to perform extended IV/2SLS,
GMM and AC/HAC, LIML and k-class regression for panel data models.
http://ideas.repec.org/c/boc/bocode/s456501.html.
15

File (1)

Content uploaded by Masataka Harada
Author content
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.