Economic Inequality and Social Capital

The Postcode Lottery runs a weekly lottery, the Street Prize, in which they randomly select a postcode in the Netherlands. All lottery subscribers in this winning postcode earn a substantial amount. This lottery generates an exogenous increase in income inequality in the neighborhood of the Street-Prize winning postcode. The Postcode Lottery also randomly selects postcodes that are awarded consolation prizes. Because these consolation prizes are small, they have a negligible effect on income inequality in neighborhoods that contain the consolation-prize winning postcode. The treatment group will comprise households living adjacent to (but outside) a postcode that won the Street Prize. The control group will comprise households living adjacent to (but outside) a postcode that won a consolation prize.

The Postcode Lottery is a popular subscription lottery in the Netherlands (www.postcodeloterij.nl). Players pay a monthly fee for a lottery subscription and are entered in all draws. A unique feature of this lottery is that, rather than drawing winning numbers, it randomly selects winning postcodes. If a postcode is selected, all lottery subscribers living in this winning postcode earn prizes.
The Postcode Lottery runs several different draws in the Netherlands. One of the most popular is the “Street Prize,” which randomly selects a postcode in the Netherlands every week. Lottery subscribers in this winning postcode earn up to €75,000, depending on the type of subscription they have. Consequently, there may be multiple households – all geographically concentrated – earning up to €75,000 each.
The Street Prize generates an exogenous increase in income inequality in the neighborhood. While the incomes of the lottery subscribers living in the winning postcode increase substantially, the incomes of households living in a different postcode, just outside the winning postcode remain the same. The latter experience some of their neighbors getting richer, but their own financial circumstances did not change.
The Postcode Lottery also randomly selects postcodes that are awarded consolation prizes. Because these consolation prizes are small, they have a negligible effect on income inequality in the neighborhoods containing the consolation-prize winning postcode.
The treatment group will comprise households living adjacent to (but outside) a postcode that won the Street Prize. The control group will comprise households living adjacent to (but outside) a postcode that won a consolation prize.

2024-04-08

2025-04-20

Our primary outcome will be an index constructed from two outcomes: (1) a question on trust and (2) behavior in an incentivized task designed to measure honesty.

Participants will be asked the following question to measure trust: “How strongly do you agree with the following statement? Most people living in my neighborhood can be trusted.” Participants will answer using a scale from 0 (“Don’t agree at all”) to 10 (“Fully agree”).
Honest behavior will be measured using the so-called “mind game.” The screen will show a card face down. Participants will be told that there is a 50-50 chance that the card is either red or black. Then, they will be asked to guess the color of the card and make a mental note of their guesses. At this point, the screen will reveal the color of the card and participants will be asked to report which color they had guessed originally. There will be 16 trials in total. At the end of the survey, one of the 16 trials will be randomly selected for payment. A participant will earn €10 if she guessed correctly in the selected trial. As a result, the participant can increase her chances of earning €10 if she misreports her guess in a trial in which she had guessed the wrong color. Although it will not be possible to identify whether a specific participant cheated, we will be able to infer cheating at the group level by comparing the reported number of correct guesses to the theoretical benchmark under truthful reporting (i.e., 50 percent).
We will construct an index (e.g., Anderson 2008) out of the trust question and the reported number of correct guesses (as a proxy for cheating). Although the index will be our main outcome, we will also separately estimate the effects on each of the two primary outcomes.
Anderson, M. L. (2008). Multiple inference and gender differences in the effects of early intervention: A reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects. Journal of the American statistical Association, 103(484), 1481-1495.

We will collect data on secondary outcomes to explore the following potential mechanisms through which income inequality may affect social capital: (a) competitiveness; (b) neighborhood cohesiveness; (c) envy; and (d) perceived unfairness. We will also measure happiness to assess the potential side-effects of inequality on mood.

To measure competitiveness, participants will be asked the following question:
“To what extent do you agree with the following statements?
Nowadays, it seems like…
a) ...winning is not the first thing; it is the only thing.
b) …life is governed by the “survival of the fittest”.
c) ...it is a dog-eat-dog world where you have to be ruthless at times.”
They will answer this question using a scale from 0 (“Don’t agree at all”) to 10 (“Fully agree”).
Neighborhood cohesiveness will be assessed using the following two questions:
a) Our social lives in the neighborhood wax and wane, changing from time to time. Over the last few months, how has the sense of community changed in your neighborhood? Participants will answer using a scale from -5 (“Much weaker sense of community”) to +5 (“Much stronger sense of community”).
b) As you think about your interactions with the people in your neighborhood over the last few months, have they become friendlier, less friendly, or have they stayed the same? Participants will answer using a scale from -5 (“Much less friendly”) to +5 (“Much more friendly”).
Envy will be measured by asking participants how strongly they agree with the following statements:
a) When I walk through my neighborhood, I sometimes feel envious of how much money some people in my neighborhood have. Participants will answer using a scale from 0 (“Don’t agree at all”) to 10 (“Fully agree”).
b) When I walk through my neighborhood, I often catch myself wishing I had as much money as some people in my neighborhood have. Participants will answer using a scale from 0 (“Don’t agree at all”) to 10 (“Fully agree”).
To measure perceived unfairness, participants will be asked how strongly they agree with the following statement:
Walking through the neighborhood, I sometimes think about how unfair it is that some people in my neighborhood have so much money while other people in my neighborhood have so little. Participants will answer using a scale from 0 (“Don’t agree at all”) to 10 (“Fully agree”).

Although the effect of the Postcode Lottery on inequality in the Netherlands as a whole is negligible, it is possible that an increase in neighborhood inequality may affect not only one’s trust in one’s neighbors but also one’s trust in people more generally. To investigate this possibility, participants will also be asked the following question: “How strongly do you agree with the following statement? Most people living in the Netherlands can be trusted.” Participants will answer this question using a scale from 0 (“Don’t agree at all”) to 10 (“Fully agree”).
Finally, the potential emotional effects of inequality will be assessed by asking participants the following question: “Overall, how happy would you say you are? Please give a number between 0 (extremely unhappy) and 10 (extremely happy)”

The Postcode Lottery also randomly selects postcodes that are awarded consolation prizes. These postcodes provide a valid counterfactual for the postcodes that win the Street Prize. Our empirical strategy will compare households living adjacent to (but outside) a postcode that won the Street Prize to households living adjacent to (but outside) a postcode that won a consolation prize.

In addition to the Street Prize, the Postcode Lottery also randomly selects postcodes that are awarded consolation prizes. For example, all lottery subscribers in one of these postcodes may earn €20.
Importantly, postcodes that win the consolation prizes are drawn from the same pool of postcodes used to select the Street-Prize-winning postcode (this pool is formed by all postcodes in the Netherlands in which there is at least one lottery subscriber). For this reason, the postcodes that win consolation prizes provide a valid counterfactual for the postcodes selected to win the Street Prize.
Our research design focuses on households living in the postcodes near to postcodes selected for either the Street Prize or a consolation prize. In particular, let “treatment households” be the households living adjacent to (but outside) a postcode that won the Street Prize. These households experience some of their neighbors getting richer, but their own financial circumstances do not change. Let “control households” be the households living adjacent to (but outside) a postcode that won a consolation prize. Because these consolation prizes are small, they have a negligible effect on income inequality in the neighborhoods of control households.
Our empirical strategy will compare treatment households to control households. Note that both control and treatment households live in non-winning postcodes, such that—irrespective of whether they have a lottery subscription or not—they will not have won any lottery prizes.

Public lottery.

The unit of randomization is a postcode.

Yes

1,134 postcodes (this is the number of targeted postcodes; in practice there will be fewer postcodes in the data because in some of them none of the invited participants will answer the survey).

We are planning to recruit subjects for about 54 weeks. Based on a response rate of 20% we expect a total sample size of roughly 1,300 participants.

To increase the number of potential participants per week, we will over-sample the control group by a factor of 2:1.

Assuming an intra-cluster correlation of ρ = 0.15, the minimum detectable effect size is approximately 0.225 standard deviations (at the 5% significance level and with a power of 80%).