Poverty and Cognitive Function

We have adapted the poverty primes by Mani et al. (2013) to the MTurk environment. As in Mani et al. (2013), we present our respondents hypothetical scenarios, each of which describes a financial problem. We randomly assign our respondents to either a hard or an easy financial scenario.

In the first financial scenario they need to explain how they would deal with an income decrease of 20% (5%) in the hard (easy) financial scenario. We then ask them a variety of questions on whether this income shock would substantially affect their situation and what kind of sacrifices they would need to make. In the second scenario people explain how they would deal with a situation in which they need to come up with an amount of money: In the hard (easy) financial scenario respondents are asked how they would come up with $3000 ($150) in a short notice. The order with which these financial scenarios is presented is randomized. Respondents write down how they might deal with the financial scenarios. The aim of exposure to these scenarios is to trigger feelings of poverty.

We have made two main changes to the primes used by Mani et al. (2013): first, we slightly increased the amounts for the hard financial scenarios. Second, we removed two financial scenarios because they did not seem well-suited for the MTurk population. We have conducted a pilot study with a sample of 350 participants on August 1st in which we document that our two primes successfully affect financial worries. In particular, poorer individuals from our sample are quite strongly affected by our treatment: They display substantially stronger financial worries and lower levels of satisfaction with income. The primes are further explained in Appendix A. Moreover, at the very end of the document we attach the exact experimental instructions.

2015-08-10

2015-08-11

- Stroop task: Our participants will be presented with a series of color words (blue, yellow, green, red). These words will appear in different colors, sometimes matching the word (e.g., the word blue, written in blue), and sometimes not matching the word (e.g., the word blue, written in yellow). The respondent's task is to indicate, as quickly and accurately as possible, the color in which the word is written, whether or not that matches the word itself. They are supposed to click the letter on the keyboard that matches the first letter of the color of the word. We incentivize the responses to this task by giving 1 cent for each correct answer, but reducing participants payoffs by 2 cents per 10 seconds it takes them to complete all items in the Stroop task and their payoff from this task cannot go negative. Respondents must complete 75 items without any time limits. In this task, we measure the number of correct answers and reaction time. We are interested in the number of correct responses and the response time to the incongruent items (where meaning of word and color do not match).

- Raven Progressive Matrices: This task measures fluid intelligence. Each trial consists of a pattern, with part of the pattern missing. Respondents are asked to choose the correct figure, from a set of 8 candidate figures, which best completes the overall pattern. We incentivize the responses to this task by giving five cents for each correct answer. Respondents must complete ten questions without any time limits. In this task, we measure the number of correct answers and reaction time. We have chosen Raven matrices that were neither too easy nor too hard based on a Pilot of 50 respondents from MTurk on August 4th.

- Financial Worries: This 4-item questionnaire provides an additional manipulation check for our poverty primes. We ask respondents to self-report on a Likert scale how worried they are about their financial situation.

- Satisfaction with income

- MacArthur Socioeconomic Ladder

We have adapted the poverty primes by Mani et al. (2013) to the MTurk environment. As in Mani et al. (2013), we present our respondents hypothetical scenarios, each of which describes a financial problem. We randomly assign our respondents to either a hard or an easy financial scenario.

In the first financial scenario they need to explain how they would deal with an income decrease of 20% (5%) in the hard (easy) financial scenario. We then ask them a variety of questions on whether this income shock would substantially affect their situation and what kind of sacrifices they would need to make. In the second scenario people explain how they would deal with a situation in which they need to come up with an amount of money: In the hard (easy) financial scenario respondents are asked how they would come up with $3000 ($150) in a short notice. The order with which these financial scenarios is presented is randomized. Respondents write down how they might deal with the financial scenarios. The aim of exposure to these scenarios is to trigger feelings of poverty.

We have made two main changes to the primes used by Mani et al. (2013): first, we increased the amounts for the hard financial scenarios. Second, we removed two financial scenarios because they did not seem well-suited for the MTurk population. We have conducted a pilot study with a sample of 350 participants on August 1st in which we document that our two primes successfully affect financial worries. In particular, poorer individuals from our sample are quite strongly affected by our treatment: They display substantially stronger financial worries and lower levels of satisfaction with income.

Randomization is done using a computer.

Individual.

No

506 respondents on Amazon Mechanical Turk.

506 respondents on Amazon Mechanical Turk.

253 treatment individuals;
253 control individuals.

The chosen sample size of 506 participants for the experiment ensures that we can detect an effect size of 0.25 at a significance level of 0.05 with a power of 0.8. Given that the effect sizes reported by Mani et al. (2013) are between 0.8 of a standard deviation and 1 standard deviation, we can be confident that our sample is sufficiently large to provide us with sufficient statistical power to detect effects.