The Impact of Menstrual Cycle Phase on Economic Choice and Rationality

Abstract

It is well known that hormones affect both brain and behavior, but less is known about the extent to which hormones affect economic decision-making. Numerous studies demonstrate gender differences in attitudes to risk and loss in financial decision-making, often finding that women are more loss and risk averse than men. It is unclear what drives these effects and whether cyclically varying hormonal differences between men and women contribute to differences in economic preferences. We focus here on how economic rationality and preferences change as a function of menstrual cycle phase in women. We tested adherence to the Generalized Axiom of Revealed Preference (GARP), the standard test of economic rationality. If choices satisfy GARP then there exists a well-behaved utility function that the subject's decisions maximize. We also examined whether risk attitudes and loss aversion change as a function of cycle phase. We found that, despite large fluctuations in hormone levels, women are as technically rational in their choice behavior as their male counterparts at all phases of the menstrual cycle. However, women are more likely to choose risky options that can lead to potential losses while ovulating; during ovulation women are less loss averse than men and therefore more economically rational than men in this regard. These findings may have market-level implications: ovulating women more effectively maximize expected value than do other groups.

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Fig 1. Example trials in the GARP and gambling experiments.

(A) Example trial in the GARP experiment. This choice set offers three bundles. The first bundle contains no cookies and 12 ounces of milk. The second bundle contains one cookie and six ounces of milk. The third bundle contains no milk and two cookies. The subject indicates their choice by checking the appropriate box with a computer mouse. (B) An example gain-only trial (left) and an example mixed gamble trial (right) in the gambling experiment. Subjects indicated their choice by button press. See Methods for more details.

Fig 2. Results from the GARP experiment.

(A) Budget sets used in the GARP experiment. Each line represents one budget set and circles on the line represent the bundles (or options) amongst which the subject can make a selection. (B) Rationality across the menstrual cycle. Measures of the Afriat’s Efficiency Index are plotted for subjects in each menstrual cycle phase and for an age-matched male control population (both n = 36 subjects). Measures for all phases were above 0.95, the common threshold for rationality, and not significantly differently across phases or compared to males (all p>0.3). Error bars represent standard errors. The dotted lines represent the AEI measurements for a random chooser (green) second grader (pink) and undergraduate (cyan) from Harbaugh et al. [22].

Fig 3. Hormone levels in the subject population.

(A) Average estradiol levels in 33 subjects, all with blood samples in all four phases. Differences between adjacent phases are significant at p<0.01 except between ovulation and the luteal phase (p = 0.063). (B) Average progesterone levels. All differences between adjacent phases are significant at p<0.01 except between menses and the mid-follicular phase (p = 0.52) and between the mid-follicular phase and ovulation (p = 0.077).

Fig 4. Results for gain-only trials in gambling experiment.

(A) Effect of menstrual phase on the number of times subject chose the safe/certain option over the risky option in gain-only trials. (B) Histogram of the number of times the safe option was chosen by each individual subject, computed as the difference between the number of safe choices in each cycle phase and the average number of safe choices across cycle phases. Fig 4A is superimposed on the histogram for comparison. (C) Effect of menstrual phase on risk aversion. Parameter fits are estimated simultaneously for all phases with controlling for session order effects (see S1 File for details on the fitting procedure). Parameter estimates for an age-matched male population are also plotted on this for comparison. Subjects were more risk averse during the luteal phase than during ovulation (p = 0.048) or the mid-follicular phase (p = 0.056).

Fig 5. Session order effects.

(A) Session order effects on loss aversion. Loss aversion increased from the first to second session, but remained stable for all subsequent sessions. Note that session order was randomized with regard to cycle phase. (B) Session order effects on risk aversion. Risk aversion increased from the first to second session, but remained stable for subsequent sessions.

Fig 6. Results from the gambling experiment for loss aversion.

(A) Effect of menstrual cycle phase on the number of times the safe certain option ($0) was chosen over the mixed gamble (ANOVA, p<0.01). (B) Histogram of the number of safe options chosen by each individual subject, computed as the difference for each subject between the number of safe options in each cycle phase and the average number of safe options chosen across cycle phases. Fig 6A is superimposed on the histogram to display group average and s.e.m. (C) Effect of menstrual cycle phase on loss aversion. Parameter fits are estimated simultaneously for all phases with the single best noise parameter (0.94 ± 0.04) controlling for session order effects. Parameter estimates for an age-matched male population are also plotted. Subjects are less loss averse during ovulation than other phases (all p<0.001). Loss aversion in menses, mid-follicular and luteal phases did not differ significantly from each other (all p>0.3).