The low expectation group could be driven by not wanting to have a child or experience of infertility. However, among people who reported having problems getting pregnant, the expected fertility variable had a higher mean than among those who did not report it. This implies that the expectation measure is largely capturing pregnancy intention. Income and wealth are highly correlated with risk preferences. Although much literature in economics assumes, a priori, that risk preferences precede income, empirical research reports evidence for both causal directions. I, therefore, exclude poverty as a control in the main models because it may be on the causal pathway between risk preferences and sexual and reproductive behavior. I, however, include a measure of poverty in the sensitivity analyses. Demand for insurance is correlated with risk tolerance, and at times itself used as a measure or risk tolerance . Therefore, I treat insurance status in the same manner as poverty, excluding it from the main models and including it for sensitivity analyses. Insurance status is coded as any insurance versus no insurance at the time of interview. The risk preference measure was missing in 3.1% of 3,821 women interviewed in 2010 or 2011. Of the 118 missing responses, 103 answered don’t know to the first gamble question. Of the 3,703 women with risk preferences measures, 530 were excluded due to reporting no sexual behavior with a member of the opposite sex in the preceding year. An additional 4% of the 3,173 responses were excluded from the analyses due to missing all of the sexual and contraceptive behavior outcomes, leaving a final analytic sample of 3,045 women. The distribution of risk tolerance for the study sample included 56.6% in the very strongly risk averse category, 23.0% in the strongly risk averse category, 10.0% in the moderately risk averse category and 10.4% is the weakly risk averse category.

These estimates parallel those reported in the literature,cannabis grow kit with one third to one half of participants falling into the most risk averse category.Table 1a shows the distribution of risk aversion by covariates in the sample. Measures of risk aversion appear similar across non-Hispanic white, non-Hispanic black and Hispanic respondents, with non-Hispanic blacks displaying slightly more responses in both risk averse and risk tolerant categories. The low numbers in the mixed/other racial group prevent any meaningful comparisons. The age range in the study sample included 25 to 31 year olds at the time of interview. With the exclusion of these 25 and 31 year olds that had low numbers of respondents, the risk preference measures appear largely stable over the age range. Higher education corresponded to decreased risk aversion and increased risk tolerance in this sample , as is consistent with the literature. As expected, risk aversion differed by marital/partnership status, with the never married/not cohabitating group showing lowest risk aversion and highest risk tolerance. Cohabitating but not married respondents also have higher risk tolerance and lower risk aversion than their married counterparts, with separated/divorces/and widowed in between. Respondents with no children ranked higher on risk tolerance and lower on risk aversion, than those with 1, 2, or 3 plus children. Insurance status, a marker for access to contraception and also SES, differed by risk aversion. Examining poverty and risk aversion also revealed expected differences. Those with lower income to poverty ratios displayed the highest degree of risk tolerance, and the wealthiest showed the highest degrees of risk aversion. Finally, risk preferences varied by expectation of future children with those that 100% expect children in the next five years, and those that 0% expect children in the next five years exhibited stronger risk aversion. Sensitivity analyses are shown in the appendices. First, I confirmed that the choice to code the second order “don’t know” responses to the middle categories of risk aversion did not alter results. All measures of effect remained unchanged. Categorizing risk preference as a linear measure of tolerance according to the adjusted mean values reported in Barsky, led to similar results . An increase in risk tolerance was associated with number of partners [RR= 1.25, 95% CI ], but not with likelihood of having a having a high-risk sexual partner.

Consistency of use results persisted for unmarried/not cohabitating women: the linear risk tolerance measure was associated with increased relative risk of inconsistent contraceptive use [RR= 3.20, 95% CI ]. The results with nonuse were attenuated from the categorical measure from the measures for the married group. Finally, for effectiveness, a one-unit increase in the linear risk tolerance measure was associated with decreased risk of using a medium effectiveness [RR=0.06, 95% CI ] or high effectiveness [RR=0.03, 95% CI ] method use versus low effectiveness methods among the unmarried/not cohabitating group. The confidence intervals for an increased relative risk of medium and highly effective method use among the married or cohabitating group contained the null. For individual methods, only the relationship with the patch or ring and sterilization persisted with the continuous risk tolerance measure. Models including the natural log of poverty and health insurance as covariates attenuated results for number of sexual partners . The consistency of contraception and effectiveness of method relationships remained similar in magnitude. To my knowledge, this is the first paper to report a relationship between risk preferences and contraceptive use, an important determinant of unintended pregnancy. In a nationally representative sample of women aged 25-31, I found that among unmarried and not cohabitating or separated women, greater risk tolerance was associated with inconsistent contraceptive use, as well as decreased use of more effective methods of contraception. Among married and cohabitating women, the highest level of risk tolerance is associated with an increased likelihood of using a more effective method of contraception. This likely indicates distinct pathways through which propensity to risk may operate in regard to reproductive behavior. I further found evidence of risk tolerance and sexual behavior, as reported number of sexual partners increased with risk tolerance. The unadjusted relationship between risk tolerance and sex with a high-risk partner did not remain after covariate adjustment. As only three percent of the sample engaged in this behavior, it may be measuring a greater level of risk taking than is captured by the four categories of risk preferences measures by the lottery question.

Despite extensive research into reproductive decisions, little attention has been paid to risk preferences. Standard economic models generally view women as rational actors who adjust fertility behavior to maximize utility. This view is inconsistent with the work from reproductive epidemiology that shows a complex set of factors influence reproductive decision making, and that women’s stated pregnancy intentions and behavior do not always converge. This paper uses a well-studied measure from economics to explain variation in reproductive risk taking. The results suggest that the propensity to assume financial risk may imply a more general risk-taking that extends to reproductive behavior. I found that partnership status modified the relationship of risk preference and contraceptive use. These results are consistent with previous work finding earlier marriage and divorce propensity are related to risk preferences.Among unpartnered women, decreasing risk aversion was associated with increased inconsistent method use. This result supports the hypothesis that more risk tolerant unmarried or separated women may accept a greater risk of pregnancy versus potential negative cost of consistently using contraception. Also less likely to use more effective methods,cannabis grow supplies this risk tolerant group may serve as a potential target of intervention. The stratified models for married or cohabitating women paint a different picture of the risk preference relationship with contraceptive use, showing that the most risk tolerant women had less use of medium and highly effective methods. This direction was the opposite of that I hypothesized. This apparent increase in method effectiveness with increasing risk tolerance may be explained by unmeasured confounding by pregnancy intention. Previous work has noted that risk tolerant women may be more likely to use less effective contraception early in life but also may be most likely to delay childbearing until later in life, despite reductions in fecundity. This latter theory may be at work in this married and cohabitating sample. The effect modification by partnership status suggests that partnership status may serve as a source of differentiation of pregnancy intention as well relationship context. If a true relationship exists between risk preferences and contraceptive behavior it may be moderated by fertility intention. Undoubtedly, inconsistent or nonuse of contraception is capturing both those women trying to become pregnant, those ambivalent to pregnancy, and those trying to avoid pregnancy. My hypothesis would operate in a different direction in the latter two groups than the former. In the absence of any measure of pregnancy planning or intention, the expectation measure may serve as a proxy for this construct. However, examining the relationship among those with low expectations of a child in 5 years did not change the results. The results by individual contraceptive methods reveal some compelling differences. The most risk averse appeared more likely to use IUDs, the most effective reversible form of contraception. Patch and ring use was higher among the most risk tolerant women, while sterilization was associated high-risk aversion. While more work on method types is needed, these results suggest that assessing women’s risk propensity may be an important determinant of appropriate contraceptive counseling. There has been a substantial push towards long acting reversible contraception for women in the US. A patient centered decision-making process could involve a discussion on only of risks of each particular method but of women’s tolerance of such risks. Indeed women may value particular features of contraceptive methods over others according to their risk preferences.

The temporal ordering of measures of risk preference and sexual and contraceptive behavior is a strength of this analysis. Economic models frequently assume that preferences are trait characteristics, and can therefore be studied at any point in time with validity. Conversely, other literature has noted fluctuation of preferences with demographic changes and life events. There is mixed evidence around potential alteration of risk preferences following childbearing. Temporal ordering may not be an issue for this analysis, however additional measures of risk preference would be beneficial to establish the preference consistency and look at contraception and pregnancy outcomes over time. The age at which risk preferences were elicited poses a substantial limitation to the theory connecting risk tolerance to fertility and contraceptive behavior. The women in this analysis were in their late twenties at the time of elicitation of risk preferences. In order to rule out problematic temporal ordering, I chose to examine only proximate contraceptive use and sexual risk behavior. However, the hypothesized relationship between risk tolerance and sexual and reproductive behavior may operate differently for women at younger and older ages than I was able to explore in this sample of 25-31 year olds. Much work in adolescent development and neuroscience shows different adaptive responses to risk . As adolescents exhibit more high-risk sexual behaviors and higher rates of unintended pregnancy, this age group would serve as a particularly important focus of further study of risk preferences and reproductive decision-making. The relationship of risk tolerance and fertility may be nonlinear with age, as Schmidt suggests. Early in fertility risk tolerant women may be less likely to use contraception while later they may be more likely to delay childbearing. Schmidt postulated that her finding that highly risk tolerant women experienced earlier first birth, was due to increased sexual risk behavior and reduced contraceptive use. While I saw no relationship with age in this sample, I may not have had enough variability in age in this time period to notice an effect. Schmidt also notes that at high education levels, risk tolerance may lead to a postponement of childbearing. While I did not see any evidence of interaction by education in this sample, additional analyses to explore subgroup differences are warranted. Over half of the sample fell into the most risk averse category. While consistent with other representative studies, this measure may miss levels of granularity in risk preference that would be important to reproductive behavior. Other questions have attempted to improve the measure to capture more heterogeneity in risk response, however, those questions were not available in the NLSY97.