do campaign contribution limits matter? evidence from the mccain-feingold act filipe campante,...
Post on 29-Dec-2015
218 Views
Preview:
TRANSCRIPT
Do Campaign Contribution Limits Matter? Evidence from the McCain-Feingold Act
Filipe Campante, Harvard Kennedy School & NBERDavid Yanagizawa-Drott, Harvard Kennedy School
Political Economy Seminar, UPFMarch 14, 2014
Motivation I
Big Picture: The role of campaign finance in a democracy
– Campaign contributions are controversial
– On the one hand, potential positive role• A healthy democracy requires political participation, and monetary contributions by
citizens is a form of participation• Can be viewed as a form of free speech• Money can mobilize voters and increase turnout – another form of political
participation• Money can improve information
– On the other hand, potential negative role• Democratic principle: “One person, one vote”• If money influences elections or policy, political influence may be disproportionally
allocated towards donors (and the wealthy in particular)
Motivation II
– Campaign finance is heavily regulated– For example, currently in the U.S.:
• Individual contributions to federal campaigns are limited to:– $2,600 for any candidate/candidate committee (per election)– $48,600 for all candidates combined, over biennial election period
• PAC contributions to federal campaigns are limited to:– Up to $5000 for any any candidate/candidate committee (per election)
– Main Question: What are the effects of limiting the amount individuals can give to candidates?
Source: Barber & McCarty (2013)
This paper
•The effect of McCain-Feingold Act of 2002• Raised the limit on individual contributions from $1000 to $2000
•We study its impact on the U.S. House of Rep.
1. Total Contributions
2. Party Balance
3. Political Outcomes– Voter mobilization– Roll-call voting
– We use an identification strategy based on how binding the limit was across donors and districts
Literature• Contribution Limits
– Surprisingly, we’ve found no study on the effects of federal limits…
– State limits (Stratmann & Aparicio-Castillo 2006, Primo & Milyo 2006, Hamm & Hogan 2008, Barber 2013)
• Limits affect: polarization, competitiveness of elections, views on government, patterns of candidacy
• Identification Problem: changes in state limits are obviously endogenous
– Theory• Coate (2004), Ashworth (2006): limits can be welfare-improving
• Campaign contributions / spending– Vast literature on effects on electoral outcomes
• Inconclusive evidence: Jacobson (1990, 2006), Levitt (1994), Gerber (1998, 2004) (Comparative survey: Scarrow 2007)
– Role of individual contributions: Ansolabehere, Figueiredo & Snyder (2003), Barber and McCarty (2013)
Outline of the Talk
• Background: McCain-Feingold Act
• Data
• Empirical Strategy: Revealed Preferences and Limits
• Results 1: Summary Stats
• Model
• Specification
• Results 2: District-Level Estimates
• Summary
McCain-Feingold Act 2002
• Limits on Individual Contributions to Candidates– Before: $1000 per candidate/election
• $25K annual total, including PACs and parties
– After: $2000 (indexed for inflation) per candidate/election• $37,500 biannual total for candidates, $57,500 for PACs and parties
• Many other provisions…– Ban on “soft money” (unregulated money spent party committees,
which could be used to affect federal elections)• Ansolabehere & Snyder (2000): soft money not very important for individual
congressional candidates
– Limits to repayment of personal loans by candidate– “Stand by your ad”: “I am co-author Filipe Campante, and I approve this
message.”– Unchanged PAC contributions
Data
– Contributions dataset (Bonica), 1990-2012• Essentially the universe of contributions• Individual donor id
– U.S. House Election outcomes• Leip Election Atlas
– Roll-Call Data• DW-Nominate First
Empirical Challenge
Empirical Challenge: Identifying the Counterfactual • What would contributions and political outcomes have been had the
limit remained $1000?• Observable 2002, fundamentally unobservable after 2002
We approach this challenge in two steps:• STEP 1: We use a simple conceptual framework (Revealed Preferences)
to calculate counterfactual contributions– Intuition: Identify donors for which a $1000 limit would bind
• STEP 2: We exploit variation in counterfactual contribution levels across Congressional Districts in 2002-04, to estimate impact on political outcomes– Intuition: Exploit that the number of donors where a $1000 limit would bind varies
Computing Aggregate Counterfactual
• We apply the conceptual framework on each individual contributor in the 2004 data– Approximately 500’000 individuals
• We calculate the counterfactual contribution in 2004, had the limit been $1000
• We define the difference as the causal effect on contributions of having a limit of $2000 instead of $1000, ceteris paribus
Total 2004 Counterfactual 2004 McCain-Feingold Effect Total 20020
50
100
150
200
250
300
The Effect of McCain-Feingold Act on Total Individual Contributions, 2004
Tota
l Ind
ivid
ual
Cont
ributi
ons,
Mill
ion
USD
District-Level Variation
– We want to identify districts where a $1000 limit would have been more binding in 2004
– We aggregate individual counterfactuals for 2004 to the district level
– Total and by party
District-Level Variation: Example
Texas District 10 Texas District 11
District-Level Variation• We define contributions due to the higher limit as:
“McCain-Feingold Contributions 2004” = (Actual 2004) – (Counterfactual 2004)
District-Level Variation: by Party
Rep. Advantage in Contributions,
2004
Counterfactual Rep. Advantage,
2004
McCain-Feingold Effect on Rep.
Advantage
Rep. Advantage in Contributions,
2002
0
5
10
15
20
25
30
The Effect of McCain-Feingold Act on Republican Advantage in Contributions, 2004
Tota
l Ind
ivid
ual
Cont
ributi
ons,
Mill
ion
USD
Effect on Republican Monetary Advantage
Model
– Motivated by these facts, we build a very simple model
• Extension of Campante (2011)
– Organize our thinking about how limits may affect:• Contributions• Parties• Voter turnout• Policy positions
Model: Assumptions
• Key ingredients– Individual contributions driven by “consumption” motive– Contributions are a normal good, but individuals can contribute only
up to an exogenously set limit» For simplicity: assume that individuals give a certain fraction of
their income, up to the limit– Voters’ policy preferences co-vary with income/wealth (e.g.
redistribution)– Two ideological parties (“D” and “R”) propose “platforms” with
commitment– Parties can use contribution money to increase turnout of their base
» E.g. Rosenstone and Hansen (1996)» Diminishing marginal returns
– Equilibrium policy is a convex combination of party platforms, with weights given by vote shares» For simplicity, to avoid discontinuities
Model
– Prediction 1: An increase in the individual contribution limit increases total individual contributions to each party
– Prediction 2: An increase in the individual contribution limit increases total turnout
• Turnout effect can be decomposed into two parts:» (I) Mobilization effect changes in how effectively parties can mobilize their base» (II) Policy effect change in the size of the parties’ bases, because of shifts in
platforms
– Prediction 3: An increase in the individual contribution limit may disproportionately benefit either party
» Eminently empirical question! » Why? It all depends on how effective the marginal dollar can be in increasing
mobilization.» General principle: If Party R (resp. Party D) gets a relative improvement in terms of
mobilization, then equilibrium policy must shift to the right (resp. to the left)
Empirical Strategy
• We exploit that our counterfactuals give us district-level variation in the infusion of money directly attributable to the McCain-Feingold Act limit increase (“McCain-Feingold Contributions”)
• Time-reversed prediction– Intuitively, this 2004 money should “disappear” in 2002– The McCain-Feingold Act is a shock that differentially hits
districts• Similar in spirit to a diff-in-diff, but in reverse
– Other idiosyncratic district shocks may of course occur during 2002-04, but this should be a good prediction on average
Empirical Strategy
Basic Identification Assumption:
Conditional on 2004 levels and 2002-04 changes in the number of contributors and unaffected contributions, 2002-04 district shocks to contributions and political outcomes are uncorrelated with “McCain-Feingold Contributions” in 2004
Example Revisited
Texas District 10 Texas District 11
Empirical Specification
• Standard Covariates: – Number of Individual Contributors (2004 Level and 2002-04 Growth)– Unaffected Individual Contributions (2004 Level and 2002-04 Growth)– State Fixed Effects
• Additional Covariates:– PAC Contributions (2004 Level and 2002-04 Growth)– Election Controls (2004 Republican Vote Share and Log Total Votes)
• Cluster Standard Errors by State
• IV-2SLS: Instrument for 02-04 Growth in Ind. Contributions by Log(MCCF)– Additional assumption: Exclusion Restriction
Reduced Form
Placebo Tests
• The identification assumption is equivalent to a “parallel trends” assumption
• To assess this, we run placebo tests for growth in outcomes across years– 2000-02– 2004-06– 2006-08
• For these years, we expect: β = 0
Reduced Form
Placebo Tests
Individual Contributions All Contributions Republican Votes Democratic Votes
00-02 04-06 06-08 00-02 04-06 06-08 00-02 04-06 06-08 00-02 04-06 06-08
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Log(McCain-Feingold Contributions) -0.007 -0.080 -0.044 -0.007 -0.063 -0.007 -0.021 -0.023 0.020 0.015 -0.017 0.025
(0.045) (0.067) (0.118) (0.041) (0.074) (0.097) (0.038) (0.020) (0.023) (0.030) (0.033) (0.017)
Covariates
Number of Contributors? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
State Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Unaffected Ind. Contributions? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Excl. Uncontested Elections No No No No No No Yes Yes Yes Yes Yes YesObservations 434 433 433 434 433 433 346 351 338 336 368 376R-squared 0.46 0.31 0.15 0.46 0.29 0.18 0.14 0.38 0.55 0.37 0.41 0.26
P-value 0.88 0.24 0.71 0.86 0.40 0.94 0.58 0.26 0.39 0.62 0.60 0.14
T1. Individual Contributions
Dependent Variable: Growth in Individual Contributions
2002-04 2002-04 2002-04 2002-04 2002-04 2002-04
(1) (2) (3) (4) (5) (6)
Log(McCain-Feingold Contributions) 0.179*** 0.179*** 0.181*** 0.178*** 0.177*** 0.167***
(0.015) (0.021) (0.022) (0.021) (0.022) (0.021)Covariates
Number of Contributors? Yes Yes Yes Yes Yes Yes
State Fixed Effects? No Yes Yes Yes Yes Yes
Unaffected Ind. Contributions? No No Yes Yes Yes Yes
PAC Contributions? No No No Yes Yes Yes
Election Controls? No No No No Yes Yes
Excluding Uncontested Elections No No No No No Yes
Observations 434 434 434 432 432 396
R-squared 0.973 0.975 0.975 0.976 0.976 0.980
T2. Total Contributions
Dependent Variable: Growth in Total Contributions
2002-04 2002-04 2002-04 2002-04 2002-04
(1) (2) (3) (4) (5)
Log(McCain-Feingold Contributions) 0.153*** 0.139*** 0.133*** 0.095*** 0.108***
(0.034) (0.038) (0.039) (0.030) (0.034)
Covariates
Number of Contributors? Yes Yes Yes Yes Yes
State Fixed Effects? No Yes Yes Yes Yes
Unaffected Ind. Contributions? No No Yes Yes Yes
PAC Contributions? No No No Yes Yes
Election Controls? No No No No Yes
Observations 434 434 434 432 432
R-squared 0.973 0.975 0.975 0.976 0.976
T3. Ind. Contributions by Party
Growth in Individual Contributions, 2002-04 Republican Candidates Democratic Candidates (1) (2) (3) (4) (5) (6) Log(McCain-Feingold Contributions) 0.177*** 0.177*** 0.162*** 0.186*** 0.177*** 0.169*** (0.054) (0.054) (0.060) (0.049) (0.051) (0.051)Covariates Number of Individual Contributors? Yes Yes Yes Yes Yes YesState Fixed Effects? Yes Yes Yes Yes Yes YesUnaffected Ind. Contributions? No Yes Yes No Yes YesPAC Contributions? No Yes Yes No Yes YesElection Controls? No No Yes No No YesObservations 336 335 335 330 329 329R-squared 0.951 0.951 0.952 0.968 0.968 0.968
T4. Voter Turnout, All Parties Growth in Number of Total Voters, All Candidates OLS OLS OLS IV-2SLS IV-2SLS IV-2SLS (1) (2) (3) (4) (5) (6) Log(McCain-Feingold Contributions) 0.0345** 0.0334** 0.0314** (0.0150) (0.0154) (0.0148) Growth in Individual Contributions, 2002-04 0.203*** 0.199** 0.188** (0.074) (0.078) (0.074)Covariates Number of Individual Contributors? Yes Yes Yes Yes Yes YesState Fixed Effects? Yes Yes Yes Yes Yes YesUnaffected Individual Contributions? Yes Yes Yes Yes Yes YesPAC Contributions? No Yes Yes No Yes YesElection Controls? No No Yes No No YesObservations 398 396 396 398 396 396R-squared 0.45 0.46 0.47 0.43 0.43 0.44
Magnitudes
– The estimates imply that a 129 000 USD increase (10%) in total contributions leads to approximately 5000 more voters
– That is, approximately 26 USD per vote
– Compared to GOTV cost-effectiveness estimates (Gerber and Green, 2008)
• Door-to-Door: 29 USD/vote• Volunteer phone calls: 38 USD/vote• Election Day Festivals: 28 USD/vote
T5. Voter Turnout, by Party Growth in Number of Voters Republican Candidates Democratic Candidates OLS OLS IV IV OLS OLS IV IV (1) (2) (3) (4) (5) (6) (7) (8) Log(McCain-Feingold Contr.) 0.0449** 0.0511*** -0.0006 -0.0027 (0.0195) (0.0182) (0.0275) (0.0288) Growth in Ind. Contributions, 02-04 0.267*** 0.314*** -0.003 -0.016 (0.101) (0.087) (0.149) (0.157)Covariates Number of Ind. Contributors? Yes Yes Yes Yes Yes Yes Yes YesState Fixed Effects? Yes Yes Yes Yes Yes Yes Yes YesUnaffected Ind. Contributions? Yes Yes Yes Yes Yes Yes Yes YesPAC Contributions? No Yes Yes Yes No Yes Yes YesElection Controls? No Yes No Yes No Yes No YesObservations 361 360 361 360 350 349 350 349R-squared 0.2878 0.3103 0.268 0.283 0.2959 0.3143 0.296 0.314
• A bit puzzling why DEM votes aren’t increasing– Even if REP get a larger boost than DEM, the latter do get more in absolute
terms
• Maybe REP use the monetary advantage to suppress DEM votes?– “Demobilization Hypothesis” (Ansolabehere et al., 1994)
• “Candidates with sufficient resources can, through the use of negative messages, keep voters away from the polls”
• Independents found most responsive to negative ads
Effects by Party
McCain-Feingold Act
Increased GOP contributions
Increase in GOP votes
Increased DEMcontributions
No increase in DEM votes
Incumbent Roll-Call Voting
Prediction 3: An increase in the individual contribution limit may disproportionately benefit either party
• General principle: If Party R (resp. Party D) gets a relative improvement in terms of mobilization, then equilibrium policy must shift to the right (resp. to the left)
– We test this using Roll Call records• DW-Nominate scores
– Timing: Incumbents facing more money in the upcoming elections
• We compare within-incumbents changes from the 107th Congress (before 2002) to 108th Congress (after 2002)
• Incumbents in both Congresses
T6. Incumbent Roll-Call Voting Conservative Voting, Change in DW-Nominate
Democrats Republicans All Incumbents
OLS OLS IV OLS OLS IV OLS OLS IV
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Log(McCain-Feingold Contr.) 0.0059** 0.0058** -0.0002 0.0002 0.0011 0.0024
(0.0025) (0.0024) (0.0041) (0.0049) (0.0021) (0.0020)
Growth in Ind. Contributions., 02-04 0.032** 0.001 0.014
(0.013) (0.033) (0.010)
Covariates
Number of Ind. Contributors? Yes Yes Yes Yes Yes Yes Yes Yes Yes
State Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes Yes
Unaffected Ind. Contributions? Yes Yes Yes Yes Yes Yes Yes Yes Yes
PAC Contributions? Yes Yes Yes Yes Yes Yes Yes Yes Yes
Election Controls? No Yes Yes No Yes Yes No Yes Yes
Observations 183 183 183 194 194 194 378 378 378
R-squared 0.31 0.33 0.28 0.29 0.29 0.29 0.19 0.30 0.29
Post 2004 Effects?
• The effects are estimated on changes in election outcomes between 2002-04
• But McCain-Feingold applies to all elections post 2002
• Since we have counterfactual contributions for all post 2002 years, we can use the estimates and extrapolate the results to later elections
Key external validity assumption:• Homogenous treatment effects across election years
Concluding Remarks
• The higher limit in McCain-Feingold Act had significant political consequences:
1. Injected large sums of money into elections• $70 Million/election (24%)• Unbalanced in favor of Republican candidates
2. Increased mobilization
3. Benefited GOP disproportionally • Shift towards more right-wing politics
Concluding Remarks
• Alternative mechanisms:– Are the effects driven entry of candidates into primaries?
• We find no evidence of entry effects, for either party
– Other potential mechanisms? (suggestions welcome)
• Mechanisms that remain to be understood
– Why wasn’t there a mobilization effect for Democrats?• “Demobilization Hypothesis”?• Currently collecting data on political ads
Thank You!
top related