Intrahousehold Belief Disagreement and Portfolio Choice

2021-07-12

2023-07-07

Both the extensive (a binary variable) and intensive (a continuous variable between 0 and 1) margins of stock market participation.

I recruit subjects on Amazon's Mechanical Turk (mTurk), an online platform that enables researchers to carry out survey experiments.

At the start of the survey, I will ask six demographic questions to restrict my sample to married individuals aged between 25 and 64.

In the first part of the survey, I will ask subjects how they think about the stock market. Specifically, I will ask them about their expected stock market returns over the next 12 months. For subjects recruited in the first batch only, I will continue to ask participants about the realized returns of the stock market in 2020 as well as in the first half of 2021. For a random subsample of subjects, I will then provide them with information about the recent performance of the stock market and ask them again about their expected stock market returns over the next 12 months.

In the second part of the survey, I will ask all subjects to imagine a scenario where they have $10,000 in a risk-free savings account that pays 1% of interest per year. For subjects recruited in the first batch, I will randomize them into three disagreement conditions: (i) with spouse having the same stock market expectation; (ii) with spouse having a more optimistic stock market expectation; and (iii) with spouse having a more pessimistic stock market participation. For subjects recruited in the second batch, I will randomize them into two sender conditions and three disagreement conditions: (i) with spouse having the same stock market expectation; (ii) with spouse having a more optimistic stock market expectation; (iii) with spouse having a more pessimistic stock market expectation; (iv) with a stranger having the same stock market expectation; (v) with a stranger having a more optimistic stock market expectation; and (vi) with a stranger having a more pessimistic stock market expectation. I will then ask all subjects to make the decision for their family on how much, if any, they would take out of the savings account to buy and hold a stock market index for a year.

For my analysis, I will first examine whether disagreement in stock market expectations between spouses is reflected in household portfolio choice using all the subjects who are informed about their spouses' beliefs (rather than a stranger's belief). In particular, I hypothesize that everything else equal, individual are more (less) likely to participate in the stock market if their spouse is more optimistic (pessimistic) about future stock returns, in both the extensive and intensive margins.

I will then limit the sample to subjects recruited in the first batch and investigate the role of information in the intrahousehold financial decision-making process. In particular, I will test whether the impact of intrahousehold belief disagreement becomes weaker if the decision maker has more information about the stock market.

I will also examine heterogeneity by gender. For example, women might incorporate more of their husband's expectation in their decision-making process. Similarly, information about the stock market might play a more important role for women.

Finally, I will limit the sample to subjects recruited in the second batch and examine whether receiving a pessimistic (or optimistic) signal from a spouse differs significantly from receiving the same message from a stranger.

For the first batch, Qualtrics will randomize respondents into one of the two information conditions as well as one of the three disagreement conditions. For the second batch, Qualtrics will randomize respondents into one of the two sender conditions as well as one of the three disagreement conditions.

Individual

No

3,000 married individuals aged between 25 and 64 for the first batch in 2021, and another 2,000 for the second batch in 2023.

3,000 married individuals aged between 25 and 64 for the first batch in 2021, and another 2,000 for the second batch in 2023.

For the first batch, approximately 500 subjects in each of the 3 x 2 (i.e., disagreement x information) condition, totaling 3,000 individuals. Given the uncertainty of the number of married workers aged between 25 and 64 available on mTurk, I plan to recruit subjects every weekday from 9 am to 5 pm (Eastern time) starting from July 12. 2021 (i.e., the first Monday after the July 4th long weekend) till the end of August 2021. The recruiting process will end when the sample size reaches 3,000 or on August 31, 2021, whichever is earlier.

For the second batch, approximately 333 subjects in each of the 3 x 2 (i.e., disagreement x sender) condition, totaling 2,000 individuals. The same recruiting procedure as the first batch from May 30, 2023 to July 7, 2023.