Extrinsic Rewards, Intrinsic Motivation and Voting

Is rewarding voters with money valuable? “Some questioned the legality of the proposal (Crawford 2006), while others debated its merit on philosophical grounds (Archibold 2006). Scholars have also debated reforms that advocate paying people to vote (or offering some type of incentive to combat anemic turnout), arguing that such initiatives would be valuable, legitimate, and effective (Hasen 2000; Karlan 1994).” Karlan compares this to compensating jurors in some democracies - what’s the difference?”

Apparently this isn’t new:

“Early on in American history, party machines habitually distributed money, liquor, and other gifts to those who turned out, focusing the selective incentives on their loyal bases (Bensel 2004). More recently, in 1999, the California state Democratic Party infamously offered vouchers good for free chicken dinners to residents of Alameda county who mailed in ballot stubs to demonstrate they voted in the state Assembly election (Pimentel 1999). In 1995, merchants in San Ramon, California, offered people with voting stubs ‘‘savings on hams and oil changes, free checking at a local bank, restaurant discounts, and even a visit to a chiropractor’’ (quoted in Hasen 2000, 1355).”

This article discusses a practical field experiment that tested financial incentives in various amounts and methods.

Incentives and prosocial motivation

Economists and psychologists have argued over intrinsic vs extrinsic motivations. Over the past decade it seems clearer that one can come at the expense of the other and mixing them isn’t always wise. Extrinsic motivations can decrease intrinsic ones. This effect has been termed ‘crowding out’ the other motivation.

Important to note there is a difference between reducing negative incentives (like the cost to do something) vs actively rewarding with an extrinsic motivating factor. That is to say - reducing the line to vote, or paying for transportation to vote isn’t the same as giving money even if they had the same financial impact.

Voting can be driven by different motivations:

Civic duty
Influence policy
Selective interest
Entertainment
Be part of a group
Internal signaling of worth

These are all intrinsic motivations that extrinsic motivations could hurt. However, this could be a factor of motivation amount: “External incentives act as a signal of their personal (in)competence.6 Gneezy and Rustichini (2000) showed that students who were offered a small amount of money to take a quiz do worse than those who were not offered compensation, but the authors found that performance improved with larger rewards.”

Adding extrinsic, monetary, motivations add a new calculation to the mix, that is directly related - the opportunity cost of time.

Voting formulate:

“ This model suggests an individual citizen will vote if: $pB + D > C$

In this formulation, p represents the probability the individual’s vote is pivotal, B is the difference in utility the individual perceives between the candidates’ attributes or policy positions, D is the direct benefit from voting (labeled D for ‘‘civic duty’’), and C represents the costs associated with voting.

The theoretical model described above implies a number of hypotheses that have been subjected to considerable empirical scrutiny. Researchers have shown, for example, that turnout is often higher in competitive contexts (in which voters’ perceptions of being pivotal are heightened), when there are clear differences between candidates (B is greater) or when costs are lowered (Mueller 2003). Studies also show civic-duty considerations or mobilization efforts can accrue benefits that raise electoral participation (Blais, Young, and Lapp 2000; Green and Gerber 2008; Mueller 2003), but the D term is more encompassing. Recent scholarship hones in on disentangling the complicated aspects of this element in the voting calculus (Gerber, Green, and Larimer 2008), which would typically incorporate the effects of selective benefits like extrinsic rewards for voting.”

D can be disaggregated into two sources: $D = U(D_I, D_E)$ .

As such, a citizen’s utility from voting given $D_I$ and $D_E$ is a function of $D_I$ (the intrinsic benefits associated with voting, a term that captures the positive feeling the voter experiences from fulfilling a civic duty, regardless of any other consequences associated with the act) and $D_E$ (the extrinsic benefit from voting, a term which could capture any other direct consequences of voting).

In this paper he’ll explore:

U(D_I, D_E) \approx \beta_1D_I + \beta_2D_E + \beta_3D_ID_E

This adds a new factor where $D_I, D_E$ affect each other with a new $\beta$ . Unfortunately, it’s very hard to measure intrinsic motivation, so everything is an approximation.

1st Experiment:

Experiment was in California for municipal elections in Gilroy.

💡

Municipal elections might have different dynamics than others. On the one hand they are local and folks care intrinsically. On the other hand, they have a much smaller turnout than national elections, implying that less overall people care, but perhaps those who do “care more”? This could have implications on the amount that extrinsic motivations can influence. For example, hypothetically, less people care and therefore when financial incentives are presented, they need them to be meaningful enough to be worth the ‘schlep’. On the other hand some who partially care could be tipped over the fence with a smaller nudge?

5 groups of subjects: no mail, mail to vote, mail with amounts of money who were told the sponsors recognized voters incur costs to vote, and they were being offered a financial reward for compliance. The level of financial reward was randomly varied to be $2, $10, or $25.

💡

Framing here is to negate opportunity cost, not ‘upside’. It’s about reducing the downside you have in voting, not rewarding you for good behavior.

To receive the reward they had to jump through some hoops: “To receive the reward, subjects were required to sign and return the postcard by a designated date. They were told they would receive the reward once participation was verified using official voter records (see Online Appendix B for mailing samples.) The mailing was arranged for subjects to receive the postcards approximately three to seven days prior to the election.”

Results

Experiment 1 (2007):

Interestingly, only 5 of the possible ~650 voters actually requested payment: two for $2, two for $10 and one for $25. Perhaps the hassle wasn’t worth it (saving the postcard and mailing back). Perhaps they didn’t want to ‘sully’ their voting with a financial incentive.
Turnout (average turnout of 20.9%):

reminder postcard with no financial incentive was essentially identical to average 20.6%.
$2 reward voted at a lower rate (18.7%), suggesting this incentive may have actually depressed participation by more than 2 percentage points on average, relative to the control group.
$10 treatment condition voted at a rate of 23.0%
$25 treatment was 24.5%

Not statistically significant.

Experiment 2 (2010):

As 1st - municipal election, only larger by orders of magnitude and added the ability to be paid on location of voting.
While directionally similar, the results are inconclusive:

$2 showed no affect
$10 showed a clear boost
$25 showed a boost in both but a smaller increase in the on location than $10, and smaller than mail back.

Using linear regression we can estimate: “A $50 incentive, for example, would increase turnout by about 7.5 percentage points on average. Taking the standard error of this prediction estimate (2.6 percentage points) into account to define a 95% confidence interval suggests the average bump in turnout attributable to a $50 incentive would likely range from a low of about 2.4 percentage points to a high of about 12.8 percentage points.”

In the on location payback: “By contrast, 24.5% in the $10 incentive condition voted in the election, implying a sizeable intent-to-treat effect of 7.5 percentage points and suggesting this incentive level elevated turnout considerably. Surprisingly, turnout among subjects assigned to the $25 incentive condition was 15.5%, suggesting the $25 incentive did little to stimulate voting in this stratum.”

Discussion

💡

• Voters have a distribution of 'care'. Some really care and will vote no matter what and then the rest of the distribution cares to varying degrees. For these different incentives (intrinsic - extrinsic) will get people 'over the line'. It's important to understand the voter base to implement these. • It’s unclear from the experiment the distribution of affected voters: whether some were deterred and who they are. Perhaps ‘smarter’ or more ‘caring’ voters were put off by the extrinsic motivation. The top line numbers give no input regarding who voted. • Other reward mechanisms could be used - gambling plays on a different emotional reward mechanism than a linear or ‘certain’ reward

CAC of a voter according to this study is very high and not the most effective way to spend to increase voting. More effective ways include media buying, thanking for previous voting and knocking on doors and calling out the vote. All cheaper and more effective on a per voter basis.