Human Reproduction, Vol.31, No.5 pp. 1113 –1119, 2016
Advanced Access publication on March 9, 2016 doi:10.1093/humrep/dew031
ORIGINAL ARTICLE Reproductive epidemiology
Shared risk aversion in spontaneous and
induced abortion
Ralph Catalano 1,*, Tim A. Bruckner 2, Deborah Karasek1,
Nancy E. Adler 3, and Laust H. Mortensen 4
1
School of Public Health, University of California, Berkeley, CA 94704, USA 2Public Health & Planning, Policy and Design, University of California,
Irvine, CA 92697, USA 3Department of Psychiatry, School of Medicine, University of California, San Francisco, CA 94118, USA 4Department of
Social Medicine, University of Copenhagen, 1017 Copenhagen, Denmark
*Correspondence address. School of Public Health, University of California, Berkeley, CA 94704, USA. E-mail [email protected]
Submitted on August 27, 2015; resubmitted on December 7, 2015; accepted on December 21, 2015
study question: Does the incidence of spontaneous abortion correlate positively over conception cohorts with the incidence
of non-clinically indicated induced abortion as predicted by shared risk aversion?
summary answer: We find that the number of spontaneous and non-clinically indicated induced abortions correlates in conception
cohorts, suggesting that risk aversion affects both the conscious and non-conscious mechanisms that control parturition.
what is known already: Much literature speculates that natural selection conserved risk aversion because the trait enhanced
Darwinian fitness. Risk aversion, moreover, supposedly influences all decisions including those that individuals can and cannot report making.
We argue that these circumstances, if real, would manifest in conscious and non-conscious decisions to invest in prospective offspring, and therefore affect incidence of induced and spontaneous abortion over time.
study design, size, duration: Using data from Denmark, we test the hypothesis that monthly conception cohorts yielding unexpectedly many non-clinically indicated induced abortions also yield unexpectedly many spontaneous abortions. The 180 month test period
(January 1995 through December 2009), yielded 1 351 800 gestations including 156 780 spontaneous as well as 233 280 induced abortions
9100 of which were clinically indicated.
participants/materials, setting, methods: We use Box –Jenkins transfer functions to adjust the incidence of spontaneous
and non-clinically indicated induced abortions for autocorrelation (including seasonality), cohort size, and fetal as well as gestational anomalies
over the 180-month test period. We use cross-correlation to test our hypothesized association.
main results and the role of chance: We find a positive association between spontaneous and non-clinically indicated induced
abortions. This suggests, consistent with our theory, that mothers of conception cohorts that yielded more spontaneous abortions than expected
opted more frequently than expected for non-clinically indicated induced abortion.
limitations, reasons for caution: Limitations of our work include that even the world’s best registration system will not capture
all spontaneous abortions and that results may not generalize beyond Denmark.
wider implications of the findings: Our findings imply that abortion, intentional or ‘spontaneous,’ follows from a woman’s
estimate, made consciously or otherwise, of the costs and benefits of extending gestation given characteristics of the prospective offspring,
likely environmental circumstances at birth, and maternal resources.
study funding/competing interest(s): The Robert Wood Johnson Health and Society Scholars Program funded the research
described in this manuscript. None of the authors has any conflict of interest to declare.
Key words: abortion / environmental effects / epidemiology / pregnancy termination / pregnancy / risk aversion
Introduction
Much literature argues that natural selection has conserved mechanisms
by which women spontaneously abort fetuses that, if born, would fail to
thrive in infancy or childhood despite maternal investment (Trivers and
Willard, 1973; Stearns, 1987; Forbes, 1997; Møller, 1997; Haig, 1999;
Boklage, 2005; Almond and Edlund, 2007; Catalano, et al., 2008;
Grant and Irwin, 2009; Bruckner et al., 2015). This selection in utero
& The Author 2016. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved.
For Permissions, please email: [email protected]
1114
presumably protects a woman’s fitness by allowing her to invest in other
offspring more likely to reproduce (Forbes, 2009).
The literature positing selection in utero argues that the mechanisms
controlling the extension of gestation somehow estimate, without
maternal awareness, the likelihood that the prospective mother can
make the investment needed for the fetus, if born, to thrive in the prevailing environment (Catalano et al., 2009, 2014). If that likelihood, or
the breadth of its confidence interval, suggests a gamble too risky for
the mother, the gestation aborts spontaneously.
Women likely vary in the level of risk at which their gestations abort
(Rai and Regan, 2006), but a population of pregnant women will
exhibit a statistically expected value such as a mean or mode (Okasha,
2007). This expected value will vary over cohorts of pregnant women
because, among other reasons, the mechanisms that select women
into pregnancy do not draw randomly over time from all women of reproductive age (Darrow et al., 2009; Ananat et al., 2013). The incidence
of spontaneous abortion would, therefore, vary over time not only with
characteristics of the fetal population and with environmental threats to
maternal resources and infant survival, but also with mean or modal risk
aversion in the population of pregnant women.
A separate literature argues that tendencies humans exhibit when
making conscious choices arose, at least in part, from natural selection
(Hirshleifer, 1977; Robson, 1996; Hintze et al., 2015). Risk aversion,
for example, presumably became characteristic of humans because it
protected fitness in small populations (Robson, 1996; Zhang et al.,
2014). This work notes that conserved risk aversion must arise from
decisional biology broader than that which humans can report using,
not only because our non-conscious choices appear risk averse, but
also because animals not presumed to make conscious choices also
appear risk averse (Harder and Real, 1987; Lu and Perrigne, 2008;
Zhang et al., 2014).
The two lines of work noted above suggest, particularly when combined with empirical and logical rejection of the Cartesian dichotomy
(Ryle, 1949; Koestler, 1967; Damasio, 1994; Jeannerod, 1997), that selection in utero may well share risk aversion with conscious investment
decisions made in everyday life. If this were true, a population whose conscious investment decisions signal increasing risk aversion would exhibit
indicators of increased selection in utero.
Analyses of Swedish birthweight data yield findings consistent with the
above inference (Catalano et al., 2014). The authors argue that very low
weight births result, at least in part, from advances in modern medicine
that convert what would have been, until roughly the past quarter
century (Jorgensen, 2010), spontaneous abortions into live births.
They find that, as predicted, the incidence of very low weight birth
increases when the population reports relatively great aversion to assuming consumer debt (Catalano et al., 2014).
The Swedish results, although consistent with shared risk aversion,
depend on self-reported intentions to consume as a measure of risk aversion, and assume that very low birthweight gauges spontaneous abortion. Here we offer a more direct test of the proposition that cohorts
of pregnant women exhibit similar risk aversion in their conscious and
non-conscious choices to invest in offspring. We test the hypothesis
that the incidence of induced abortion, which evidence suggests
follows at least in part from conscious estimates of costs and benefits
(Smetana, 1981; Henshaw, 1995; Foster et al., 2012) and, therefore,
reflects maternal preferences such as risk aversion, will correlate positively over conception cohorts with the incidence of spontaneous
Catalano et al.
abortion. We test this hypothesis with data from Danish registries considered among the most complete and accurate sources of information
on the course of gestation (Lynge et al., 2011).
Materials and Methods
Data
We test our hypothesis with data retrieved from the Danish National Patient
Register and the Danish Medical Birth Register. We extracted monthly
counts of live births, induced abortions, and spontaneous abortions that
involve contact with a hospital or outpatient clinic (Lynge et al., 2011). This
research did not involve human or animal subjects and the data were anonymized. The Danish Data Protection Agency approved our use of these data
(#2013-41-2399) as did Statistics Denmark.
A correlation between spontaneous and induced abortions over conception cohorts could arise not only from shared variation in maternal risk aversion, but also from variation in gestational anomalies that would induce
both induced and spontaneous abortion (Neuhauser and Krackow, 2007).
Temporal variation in these anomalies could reflect endogenous stochastic
processes or maternal exposure to exogenous teratogens that damage
fetuses thereby leading to both induced and spontaneous abortions. To
help control for these anomalies, we separated induced abortions into
those with and without clinical indication of fetal or maternal complications
and used the former as a covariate in our tests. We classified non-clinically
indicated abortions as those with ICD-10 code O.04 (i.e. induced medical
abortion) and occurring before the 12th week of gestation. The Danish
Health Act makes non-clinically indicated abortions free and publicly available
before the 12th completed week of gestation. We classified clinically indicated abortions as those occurring after the 12th week of gestation and
with ICD procedure codes O.05 and O.07, in which law permits abortions
when the mother’s health is at risk or the fetus shows a gross anomaly.
We started our series with data from January 1995 when the Registers
expanded to include all in- and out-patient facilities in Denmark. Our
series ends 180 months later (i.e. December 2009) with the most recent
data available at the time of our analyses. We used gestational age at time
of birth or abortion to construct 180 monthly conception cohorts.
For live births, gestational age was missing in 0.63% of cases, for induced
abortions in 0.03% of cases, and for spontaneous abortions in 3.12% of
cases. We performed single imputation of the missing gestational ages by randomly sampling from the observed gestational ages among pregnancies with
the same outcome.
The Danish registry identified 1 351 800 gestations over the 180-month
test period. Of these, 961 560 or about 71%, yielded live births. The 156
780 spontaneous abortions represented 11.6% of registered gestations.
Clinically indicated induced abortions represented 9100 or less than 1% of
the gestations. Non-clinically indicated induced abortions (224 280 over
the test period) accounted for 16.6% of registered gestations.
Statistical analyses
Consistent with Galton’s and Pearson’s (Stigler, 1989) definition of empirical
correlation, we infer support for our theory if the observed monthly number
of spontaneous and non-clinically indicated induced abortions differ similarly
from their statistically expected values in the same conception cohorts. We
derived the expected values of spontaneous and non-clinically indicated
induced abortions through the following steps. First, we separately regressed
the monthly number of spontaneous and non-clinically indicated abortions
on the monthly number of live births and clinically indicated induced abortions. We specified the number of live births from each cohort as a surrogate
for the size of the population at risk because we cannot know the true population at risk of abortion in a conception cohort. We specified the number of
1115
Risk aversion in spontaneous and induced abortion
clinically indicated induced abortions to control, as discussed above, for frank
pathologies in fetuses and gestations.
Second, we used Box – Jenkins methods (Box et al., 2008) to model autocorrelation (i.e. trends, cycles (including seasonality), and the tendency to
remain elevated or depressed after high of low values) in the residuals of
the two regressions estimated in step 1. These well-developed methods,
which appear widely in the epidemiologic literature (Catalano and Serxner,
1987; Zeger et al., 2006), mathematically express various filters through
which time series without patterns can pass. Each filter, or combination
of filters, imposes a unique pattern of autocorrelation. The Box– Jenkins
approach uses an iterative model-building process by which the researcher
infers the filter that imposed the observed pattern.
Together, the above steps yield equations, referred to as Box– Jenkins
transfer function models, of which the fitted values become the expected
values of the dependent variable. The residuals from these equations
measure the degree to which the observed values differed from expected.
The general form of the final transfer function models for spontaneous and
non-clinically indicated induced abortions is as follows:
At = c + vXt + v2 X2t + (1 − uB)(1 − uBq )/(1 − fB)(1 − fBp )at
At is the number of spontaneous or non-clinically indicated induced abortions attributed to the Danish gestational cohort conceived in month t. c is
a constant. Xt is the number of live births attributed to the cohort conceived
in month t. X2t is the number of clinically indicated induced abortions attributed to the cohort conceived in month t. v and v2 express the unique association (i.e. net of the other covariate and autocorrelation) between
spontaneous abortions and the covariates Xt and X2t. u is a moving average
parameter that measures the tendency of perturbations to be present
for more than one month. f is an autoregressive parameter that measures
a series’ tendency to remain elevated or depressed after a perturbation.
B is the ‘backshift operator’ or value of the variable it conditions at month
t 2 q or at t 2 p. The error term, at, measures the difference between the
number of observed spontaneous or non-clinically indicated abortions
yielded by the cohort conceived in month t, and the number expected
from live births and clinically indicated induced abortions as well as from
any autocorrelation unique to spontaneous or non-clinically indicated abortions (i.e. not shared with live births and clinically indicated abortions).
In our third step, we tested the hypothesized association between spontaneous abortion and non-clinically indicated induced abortion by calculating
the cross-correlation function between the residuals of the two transfer functions (Catalano et al., 1983; Katz, 1988). This step computes correlation
coefficients between the two residual series in the synchronous configuration
(i.e. both series measured in the same month) as well as in the ‘lead’ (the residuals of spontaneous abortions temporally precede those for non-clinically
indicated induced abortions by 1 and 2 months) and ‘lag’ (the residuals of
spontaneous abortions temporally follow those for non-clinically indicated
induced abortions by 1 and 2 months) configurations. Results would support our theory if the synchronous coefficient exceeded twice its standard
error while the lead and lag coefficients did not.
Results
Table I shows the final transfer functions from which we estimated the
statistically expected incidence of spontaneous and non-clinically indicated induced abortions. Live births predicted, separately, both types
of abortions. Clinically indicated induced abortions predicted neither
spontaneous nor other induced abortions. The autoregressive parameters at lag 12 indicate that both spontaneous and non-clinically indicated induced abortions exhibited seasonality not shared with either
live births or clinically indicated induced abortions. The autoregressive
parameter at lag 3 shown for non-clinically indicated induced abortions
suggests that they exhibit autocorrelation in which a high or low value
at month t predicts a high or low value, although of diminishing size, 3,
but not 1 or 2, months later.
Figures 1 and 2 show the observed and the statistically expected
(i.e. fitted values from the two transfer functions) number of spontaneous and non-clinically indicated induced abortions for the conceptions
cohorts. Table II shows the cross-correlations of the two residual series.
Results support our theory in that the synchronous coefficient (i.e. 0.32)
exceeds twice its standard error (i.e. 0.08), while the lead and lag coefficients do not. These results suggest that mothers of conception cohorts
that yielded more spontaneous abortions than expected from the size of
the cohort, the frequency of frank fetal and gestational pathology in the
cohort, and history (i.e. autocorrelation), also opted more frequently
than expected for non-clinically indicated induced abortion.
To visually convey this association, Fig. 3 shows the scatter plot and
best fitting line of the residuals of the transfer functions for spontaneous
abortions and non-clinically indicated induced abortions. Table III offers
an alternative display of concordance between the two types of abortions. This cross tabulation arrays monthly conception cohorts by
whether their yield of the two types of abortions fell below or above
expected values. The disproportionate concentration of cohorts in the
upper left (i.e. both types of abortion below expected values) and
lower right (i.e. both types above expected values) cells implies convergence of the two series. The table also allows a x 2 test of independence
between the 2 types of abortions. Results support our hypothesis of convergence (x 2 ¼ 6.4896; 1 degree of freedom).
Table I Estimated parameters for transfer function models of spontaneous and non-clinically indicated induced abortions
attributed to 180 monthly (i.e. January 1995 through December 2009) Danish conception cohorts.
Estimated parameter
Abortion type
...................................................................................................................
Spontaneous
Non-clinically indicated induced
.............................................................................................................................................................................................
Constant
Live births
Clinically indicated abortions
Box Jenkins autoregressive parameters
199.9510 (101.8622; 298.0398)
0.0907 (0.0694; 0.1120)
20.5406 (21.1156; 0.0344)
B 12 ¼ 0.2439 (0.1099; 0.3779)
2141.0182 (2298.6114; 16.5750)
0.0847 (0.0530; 1164)
0.2142 (20.7697; 1.1981)
B 3 ¼ 0.2744 (0.1468; 0.4020)
B 12 ¼ 0.6295 (0.5341; 0.7249)
Superscripts indicate that abortions in any month predict those 3 or 12 months later. 95% confidence interval (2-tailed test) shown in parentheses.
1116
Catalano et al.
Figure 1 Observed (January 1995 through December 2009) and expected (January 1996 through December 2009; 12 months lost to modeling)
spontaneous abortions attributed to monthly Danish conceptions cohorts.
Figure 2 Observed (January 1995 through December 2009) and expected (April 1996 through December 2009; 15 months lost to modeling)
non-clinically indicated induced abortions attributed to monthly Danish conceptions cohorts.
We applied the outlier identification and control strategies of Chang,
Tiao, and Chen to each transfer function to remove the influence of very
large and very small values on the estimation of expected values (Chang
et al., 1988). We then repeated our cross-correlation test to determine
whether our original results arose solely from the concordance of outlying values. Results remained the same in that the synchronous coefficient
(i.e. 0.24) exceeded twice its standard error (i.e. 0.08) but the lead and
lag coefficients did not.
As shown in Table I, the coefficient for clinically indicated induced
abortions did not exceed twice its standard error in either transfer function. The constant for non-clinically indicated induced abortions did not,
moreover, exceed twice its standard error. We deleted these predictors
from the transfer functions and estimated the cross-correlations again.
The results did not change in that the synchronous coefficient remained
0.32 and its standard error remained 0.08. The lead and lag coefficients
did not exceed twice their standard errors.
1117
Risk aversion in spontaneous and induced abortion
Table II Cross-correlation coefficients of the residuals from transfer functions for spontaneous and non-clinically indicated
induced abortions.
Both series in same month
Spontaneous precedes non-clinically
indicated induced abortions by months
shown
Spontaneous follows non-clinically
indicated induced abortions by months
shown
.............................................................................................................................................................................................
2 months
1 month
Synchronous
1 month
2 months
20.08 (20.23; 0.07;)
0.01 (214; 0.16)
0.32 (0.16; 0.48)
0.05 (20.1; 0.2)
0.09 (20.06; 0.24)
95% confidence interval (2-tailed test) shown in parentheses.
Figure 3 Scatter plot and best fitting line of residuals from transfer functions of spontaneous and non-clinically indicated induced abortions yielded by 165
month Danish conception cohorts (i.e. April 1996 to December 2009).
Table III Cross tabulation of spontaneous and
non-clinically indicated induced abortions above and
below expected values for 165 Danish conception
cohorts (15 cohorts lost to modeling).
Non-clinically indicated
induced abortions
.......................................
Below
expected
Above
expected
........................................................................................
Spontaneous
abortions
Below expected
Above expected
54
31
35
45
Discussion
Spontaneous and induced abortions remain rarely studied as related indicators of risk aversion in the population (Ahmed and Ray, 2014). Our
results suggest that estimates, made consciously and non-consciously,
of the rewards and costs of pregnancy influence both classes of abortion.
This circumstance further suggests that the scholarship intended to
describe how and why we make conscious choices in general (Kahneman
and Tversky, 1979; Kahneman, 2011), and reproductive choices in particular (Becker, 1960; Hass, 1974; Blum and Resnick, 1982; Hollerbach,
1983; Barro and Becker, 1989; Foster et al., 2012), may apply to decisions
we cannot report or describe making. If so, behavioral economics and
portfolio theory (Forbes, 2009) may add to our understanding of spontaneous abortion and the timing of parturition.
Our analyses benefit from the unusually complete and detailed Danish
registries of gestation. The quality of these data allows us to apply welldeveloped and widely accepted time-series routines. These methods
rule out, for example, that our findings could arise from seasonality or
any other shared autocorrelation or from third variables that also
affect the incidence of clinically indicated induced abortions such as
chromosomal abnormalities or exposure to teratogenic stressors.
Although the high quality of the Danish registry likely increases the internal validity of our test, the characteristics of Danish culture that lead to
thorough documentation of gestation may reduce the external validity of
our findings. Danish women do not, for example, pay fees for induced
abortions perhaps making the procedure more common in Denmark
1118
than in countries with less liberal financing. Replication could establish
how widely our results apply but data from other societies may be of
lesser quality and therefore reduce confidence in findings.
As with all observational studies, readers will apply the law of parsimony
in choosing between our theory and its rivals including, for example, the
intentional misclassification of induced abortions as spontaneous. Intentional misclassification could explain our findings if women who feared
the stigma of induced abortion either deceived trained medical staff into
diagnosing an induced abortion as spontaneous, or recruited staff into
complicit deception. We know of no evidence of either circumstance in
Denmark. Indeed, Danes use the National Patient Register, among
other purposes, to assess and ensure the quality of medical services
thereby making accurate reporting in the interest of medical staff.
We argue that abortion, intentional or ‘spontaneous,’ follows from a
woman’s estimate, made consciously or otherwise, of fetal Darwinian
fitness given characteristics of the prospective offspring, likely environmental circumstances at birth, and maternal resources. We anticipate
that our invocation of non-conscious maternal estimates of costs and
benefits may lead readers to seek more conventional explanations of
our findings. We note, therefore, that research has shown humans
quite capable of non-conscious estimations and choices (Arrow, 1965;
Bateson, 2007; Beaumont et al., 2009). Our argument, furthermore,
arises from a principle widely accepted in evolutionary theory—that
decisional biology, whether or not cognitively available, arises from
natural selection and therefore operates, at least in part, to protect
fitness (Hirshleifer, 1977; Robson, 1996; Hintze, et al., 2015).
Our findings have implications for the applied literature concerned with
‘fetal origins’ of morbidity (Barker, 1995; Calkins and Devaskar, 2011). Although stressful times may, as the fetal origins literature posits, ‘dysregulate’ some aspects of gestation, not all the morbidity expected from
such circumstances will appear in birth cohorts because both spontaneous
and induced abortion could ‘cull’ less fit fetuses (Catalano and Bruckner,
2006). Our data do not allow us to test whether cohorts that yielded
more spontaneous and induced abortion also yielded infant morbidity different from that of other cohorts. Researchers could, however, test this
possibility by linking data like ours to health records.
Our findings also contribute to the basic literature by suggesting that
the non-conscious decisional biology of gestation follows rules similar
to those that guide consciously made investment decisions (Catalano
et al., 2014). These rules presumably include that mothers ‘optimize’
their fitness through, among other possible mechanisms, spontaneous
abortion. The correlation we found should stimulate further inquiry
regarding the extent to which both spontaneous and induced abortion
flow from biologically conserved adaptations.
Authors’ roles
R.C. developed the research question, performed the analyses and drafted
the manuscript. T.A.B. contributed to the data acquisition and manuscript.
D.K. contributed to the interpretation of results and to the manuscript.
N.E.A. contributed to study design and the manuscript. L.H.M. acquired
the data and contributed to the method and the manuscript
Funding
The Robert Wood Johnson Health and Society Scholars Program funded
the research described in this manuscript.
Catalano et al.
Conflict of interest
None declared.
References
Ahmed S, Ray R. Determinants of pregnancy and induced and
spontaneous abortion in a jointly determined framework: evidence
from a country-wide, district-level household survey in India. J Biosoc
Sci 2014;46:480 – 517.
Almond D, Edlund L. Trivers-Willard at birth and one year: evidence from US
natality data 1983 – 2001. Proc Biol Sci 2007;274:2491– 2496.
Ananat EO, Gassman-Pines A, Gibson-Davis C. Community-wide job loss
and teenage fertility: evidence from North Carolina. Demography 2013;
50:2151 – 2171.
Arrow KJ. Aspects of the Theory of Risk-Bearing. Helsinki: Yrjö Jahnssonin
Säätiö, 1965.
Barker DJ. Fetal origins of coronary heart disease. BMJ 1995;311:171– 174.
Barro RJ, Becker GS. Fertility choice in a model of economic growth.
Econometrica 1989;57:481 – 501.
Bateson M. Recent advances in our understanding of risk-sensitive foraging
preferences. Proc Nutr Soc 2007;61:509– 516.
Beaumont HJE, Gallie J, Kost C, Ferguson GC, Rainey PB. Experimental
evolution of bet hedging. Nature 2009;462:90 – 93.
Becker GS. An economic analysis of fertility. In: Demographic and economic
change in developed countries. Columbia University Press, 1960, 209 – 240.
Blum RW, Resnick MD. Adolescent sexual decision-making: contraception,
pregnancy, abortion, motherhood. Pediatr Ann 1982;11:797 –805.
Boklage CE. The epigenetic environment: secondary sex ratio depends on
differential survival in embryogenesis. Hum Reprod 2005;20:583 – 587.
Box G, Jenkins G, Reinsel G. Time Series Analysis: Forecasting and Control, 4th
edn. Hoboken, NJ: Wiley, 2008.
Bruckner TA, Helle S, Bolund E, Lummaa V. Culled males, infant mortality and
reproductive success in a pre-industrial Finnish population. Proc Biol Sci
2015;282:20140835.
Calkins K, Devaskar SU. Fetal origins of adult disease. Curr Probl Pediatr
Adolesc Health Care 2011;41:158 – 176.
Catalano R, Bruckner T. Secondary sex ratios and male lifespan: damaged or
culled cohorts. Proc Natl Acad Sci USA 2006;103:1639 – 1643.
Catalano R, Serxner S. Time series designs of potential interest to
epidemiologists. Am J Epidemiol 1987;126:724– 731.
Catalano RA, Dooley D, Jackson R. Selecting a time-series strategy. Psychol
Bull 1983;94:506.
Catalano R, Bruckner T, Smith KR. Ambient temperature predicts sex ratios
and male longevity. Proc Natl Acad Sci USA 2008;105:2244– 2247.
Catalano RA, Saxton K, Bruckner T, Goldman S, Anderson E. A sex-specific
test of selection in utero. J Theor Biol 2009;257:475 – 479.
Catalano R, Karasek D, Gemmill A, Falconi A, Goodman J, Magganas A,
Hartig T. Very low birthweight: dysregulated gestation versus
evolutionary adaptation. Soc Sci Med 2014;108:237– 242.
Chang I, Tiao G, Chen C. Estimation of time series parameters in the
presence of outliers. Technometrics 1988;30:193– 204.
Damasio AR. Descartes’ Error : Emotion, Reason, and the Human Brain.
New York: Putnam, 1994.
Darrow LA, Strickland MJ, Klein M, Waller LA, Flanders WD, Correa A,
Marcus M, Tolbert PE. Seasonality of birth and implications for temporal
studies of preterm birth. Epidemiology 2009;20:699– 706.
Forbes LS. The evolutionary biology of spontaneous abortion in humans.
Trends Ecol Evol 1997;12:446– 450.
Forbes S. Portfolio theory and how parent birds manage investment risk.
Oikos 2009;118:1561 –1569.
Risk aversion in spontaneous and induced abortion
Foster DG, Gould H, Taylor J, Weitz TA. Attitudes and decision making
among women seeking abortions at one U.S. clinic. Perspect Sex Reprod
Health 2012;44:117 – 124.
Grant VJ, Irwin RJ. A simple model for adaptive variation in the sex ratios of
mammalian offspring. J Theor Biol 2009;258:38– 42.
Haig D. Genetic conflicts of pregnancy and childhood. In: Stearns SC (ed).
Evolution in health and disease, Oxford: Oxford University Press, 1999;
77 – 90.
Harder LD, Real LA. Why are bumble bees risk averse? Ecology 1987;
68:1104– 1108.
Hass PH. Wanted and Unwanted Pregnancies: A Fertility Decision-Making
Model. J Soc Issues 1974;30:125– 165.
Henshaw SK. Factors hindering access to abortion services. Fam Plann
Perspect 1995;27:54 – 87.
Hintze A, Olson RS, Adami C, Hertwig R. Risk sensitivity as an evolutionary
adaptation. Sci Rep 2015;5:8242.
Hirshleifer J. Economics from a biological viewpoint. J Law Econ 1977;20:
1 – 52.
Hollerbach PE. Fertility decision-making processes: a critical essay. In: Lee RD
(ed). Determinants of Fertility in Developing Countries. New York: Academic
Press, 1983, 340 – 380.
Jeannerod M. The Cognitive Neuroscience of Action. Oxford, OX, Cambridge,
MA: Blackwell, 1997.
Jorgensen M. Born in the USA– The History of Neonatology in the United
States: A Century of Caring. NICU Currents June 2010:8 – 12.
Kahneman D. Thinking, Fast and Slow. New York, NY: Farrar, Straus and
Giroux, 2011.
Kahneman D, Tversky A. Prospect theory: an analysis of decision under risk.
Econometrica 1979;47:263 – 292.
Katz RW. Use of cross correlations in the search for teleconnections.
J Climatol 1988;8:241 – 253.
1119
Koestler A. The Ghost in the Machine. London: Hutchinson, 1967.
Lu J, Perrigne I. Estimating risk aversion from ascending and sealedbid auctions: the case of timber auction data. J Appl Econ 2008;
23:871– 896.
Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register.
Scand J Public Health 2011;39:30 – 33.
Møller AP. Developmental selection against developmentally unstable
offspring and sexual selection. J Theor Biol 1997;185:415 – 422.
Neuhauser M, Krackow S. Adaptive-filtering of trisomy 21: risk of Down
syndrome depends on family size and age of previous child. Die
Naturwissenschaften 2007;94:117 –121.
Okasha S. Rational choice, risk aversion, and evolution. J Philos 2007;104:
217 – 235.
Rai R, Regan L. Recurrent miscarriage. Lancet 2006;368:601– 611.
Robson AJ. A biological basis for expected and non-expected utility. J Econ
Theory 1996;68:397 –424.
Ryle G. The Concept of Mind. London, New York: Hutchinson’s University
Library, 1949.
Smetana JG. Reasoning in the personal and moral domains: adolescent and
young adult women’s decision-making regarding abortion. J Appl Dev
Psychol 1981;2:211– 226.
Stearns SC. The selection-arena hypothesis. In: Stearns SC (ed). The Evolution of
Sex and Its Consequences. Basel: Birkhauser Verlag, 1987, 337–388.
Stigler SM. Francis Galton’s account of the invention of correlation. Stat Sci
1989;4:73 – 79.
Trivers RL, Willard DE. Natural selection of parental ability to vary sex ratio of
offspring. Science 1973;179:90 – 92.
Zeger SL, Irizarry R, Peng RD. On time series analysis of public health and
biomedical data. Annu Rev Public Health 2006;27:57 – 79.
Zhang R, Brennan TJ, Lo AW. The origin of risk aversion. Proc Natl Acad Sci
USA 2014;111:17777 – 17782.