We are happy to announce that all trial registrations will now be issued DOIs (digital object identifiers). For more information, see here.
The Introduction of Hormonal Contraception and the Incidence of Suicide Among Adolescent Girls
Last registered on November 20, 2019


Trial Information
General Information
The Introduction of Hormonal Contraception and the Incidence of Suicide Among Adolescent Girls
Initial registration date
November 19, 2019
Last updated
November 20, 2019 2:56 PM EST
Primary Investigator
Stockholm School of Economics
Other Primary Investigator(s)
Additional Trial Information
Start date
End date
Secondary IDs

Evaluate whether the introduction of hormonal contraception (HC) increased or decreased the risk of completed suicide among adolescent girls and women and, more generally, present evidence on the association between the diffusion of HC across Swedish markets and mortality by cause and age.


HC is arguably the most important scientific advance of the last century. Regarded as a safe, effective, and affordable contraceptive method it is used by millions of women around the world. Yet, recent medical research has revealed a significant association between hormonal contraceptive use and subsequent depression and suicide; an effect most pronounced among adolescents. This study aims to estimate the effect of HC on suicide and mortality in a historical setting where contraceptive alternatives are limited and abortion is illegal.


This study aims to be the first to evaluate the short-run relationship between suicide and use of hormonal contraception (HC) proximate to the legalization of the pill. Recent medical research has pointed to an association between low dose HC use, depression, and suicide; an association most pronounced among adolescents (see Skovlund et al (2016), Skovlund et al (2018)). Yet, other epidemiological studies using have found a reduction in depressive symptoms associated with HC use (see Keyes (2013)). The relationship between HC use, mental health and suicide has been understudied both in population based epidemiological studies as well as the economic literature on the social and economic impact of technological innovations such as the pill. The lack of attention is particularly striking in the Swedish setting; within a decade of legalization over a quarter of the fecund female population was using high dose oral contraception on an annual basis, among these more than half of teenage girls.

Did the fertility control effects of the pill lead to a reduction in pregnancies and suicides among young women? This is a relevant question; autopsy data illustrates how pregnancy was common among Swedish women whom committed suicide (see Hedren (1901)). Although recent studies provide evidence of a plausible causal channel between HC diffusion and increased suicide, the total effect of this contraceptive innovation on suicide and mortality has not been studied. I combine Swedish market level data on HC sales with detailed individual data on cause of death for 2.5 million women born in Sweden from 1915-1964 to quantify the short-run impact of HC diffusion on female mortality, with a focus on suicide among adolescent girls.

The introduction of the pill is a natural experiment which allows for plausible comparisons of counterfactual suicide behavior in the absence of HC, an innovation relative to recent studies which rely on women whom never use HC as controls. I will compare changes in female mortality trends before and after the legalization of HC across geographic markets and age groups. Two types of empirical models are employed. A differences-in-differences-in-differences (DDD) design compares changes in mortality before-after the pill, among high-low pill use age groups, and high-low pill use markets, providing a descriptive analysis of the correlation between HC diffusion and suicide trends. This descriptive analysis will be supplemented by an instrumental variables (IV/OLS) model to provide causal evidence on HC use and suicide. Under certain assumptions the IV estimates provide an estimate of the causal effect of HC use on suicide behavior, absent the effects of omitted variables which may skew correlations between HC and suicide, hence providing a check on the DDD analysis and the importance of omitted factors.

OLS estimates use the actual diffusion of HC to estimate correlations between HC use and suicide behavior, while IV estimates will use predictions of HC use driven by measures from a century earlier to identify exogenous shifts in HC use and in turn the causal effects of HC on suicide and mortality. The choice of instruments relies on historical demographic and economic data, such as non-marital birth rates or wage and commodity prices from the 19th Century, to capture latent demand for HC unrelated to contemporaneous shocks. These pre-determined differences in latent demand are used to identify an exogenous source of variation in HC use.

External Link(s)
Registration Citation
ragan, kelly. 2019. "The Introduction of Hormonal Contraception and the Incidence of Suicide Among Adolescent Girls." AEA RCT Registry. November 20. https://doi.org/10.1257/rct.5059-1.0.
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
The nationwide approval of hormonal contraception (HC) by the Swedish Board of Health and Welfare in 1964 and subsequent take-up of HC by women in Sweden from 1964-1974.

Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
The study evaluates the short-run effect of HC access and use on suicide among adolescent girls and women following the legalization of HC in Sweden late in 1964 through 1974. Suicide (ICD 8 Codes 950-59) per 100,000 women in specific age and location of residence cells at annual and quarterly frequencies will be the primary measure of this outcome.

Violent and accidental deaths:
A closely related measure is death by accident of undetermined intent (ICD 8 Codes 980-89) and the more general class of violent and accidental causes of death (ICD 8 Codes 800-999) per the female population of a given age and location.

Deaths during pregnancy, childbirth and the puerperium:
Deaths during pregnancy, childbirth and the puerperium (ICD 9 Codes 630-679) includes female mortality occurring as a result of abortion or other pregnancy related complications, as well as mortality during childbirth and in the period following childbirth. Mortality by cause and in total within class per 100,000 women by specific age and location of residence cells will be the primary measures of these outcomes.
Primary Outcomes (explanation)
All measures will be expressed in terms of mortality per 100,000 population in levels and the inverse hyperbolic sin to approximate the log function without dropping zeroes.
Secondary Outcomes
Secondary Outcomes (end points)
Suicide by demographic sub-groups:
Mortality rates by suicide (as described above) may be further disaggregated by marital status or other demographic characteristics such as parental education and occupation.

Violent and accidental death by demographic sub-groups:
Mortality due to violent and accidental causes (as described above) may be further disaggregated by marital status or other demographic characteristics such as parental education and occupation.

Deaths during pregnancy, childbirth and the puerperium by demographic sub-groups:
Mortality during pregnancy, childbirth and the puerperium may be further disaggregated by demographic characteristics such as marital status or parental education/occupation.

All-cause mortality and other general and specific causes identified in the medical literature:
Following the analysis of Vessey et al (1989) and Beral et al (1999) we will estimate correlations between HC diffusion and all cause mortality as well as several general and specific classes of mortality described in these studies. Following Beral et al (1999), this would include the following general groupings: all causes (ICD-8 codes 000-999), all cancers (ICD-8 140-209), all circulatory diseases (ICD-8 390-458), all digestive diseases (ICD-8 520-577), and all other diseases (ICD-8 1-139, 210-389, 460-519, 578-799), as well as the previously mentioned violent and accidental causes (ICD-8 800-999). Several specific cancers are also considered (colorectal (ICD-8 153-154), lung (ICD-8 162), breast (ICD-8 174), cervix (ICD-8 180), ovary (ICD-8 183)) as well as specific circulatory diseases (ischaemic heart disease (ICD-8 410-414), other heart disease (ICD-8 420-429), cerebrovascular disease (ICD-8 430-438)) and liver disease (ICD-8 570-573). These mortality measures may also be disaggregated by demographic characteristics to be comparable to the analysis above.
Secondary Outcomes (explanation)
All measures will be expressed in terms of mortality per 100,000 population in levels and the inverse hyperbolic sin to approximate the log function without dropping zeroes.
Experimental Design
Experimental Design
Experimental Design, Setting and Participants:

A prospective cohort study consisting of all women born in Sweden between 1915 and 1964 (2.5 million individuals). The approval of HC by the Swedish Board of Health and Welfare is a natural experiment; the suicide and mortality behavior of women before HC was legalized can be compared to the suicide behavior of subsequent cohorts. Differential take-up of HC across Swedish markets, as captured in quarterly market level HC sales data, also leads to differential exposure to HC across markets and time. Differential HC use across age groups allows for further comparison across geographic units, age groups, and time periods which experienced differential exposure to HC. Data on cause of death (ICD Codes) is provided from the Swedish Board of Health and Welfare and linked to the anonymized population registry by Statistics Sweden. The sampled population is linked with parents/siblings/children/spouses (10+ million individuals).


Individual HC use is not observed. Instead market level HC sales per the fecund female population at quarterly frequencies is the measure of HC exposure used. The high dose HC women in this study were exposed to are generally much higher in both estrogen and progestin content than the low dose HC studied in Skovlund et al (2018). Conluten, a leading brand of HC during my period of study contained 2 mg of norethisterone and 0.1 mg of mestranol. This implies an estrogen content double that of combination HC common in later studies.
Experimental Design Details
Randomization Method
Exposure to HC by market is endogenous after the legalization in 1964, prior to this HC use is uniformly zero per woman in every market.
Randomization Unit
The analysis will be clustered at the geographic level of the labor market defined by Statistics Sweden as an A-region.
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters

The definition of a market used in the Swedish Drug Market (SDM) publication from which drug sales data is collected follows the definition of a labor market by Statistics Sweden (A-region); there are 70 markets and in turn clusters according to this definition. Cluster robust standard errors will be reported throughout the analysis.
Sample size: planned number of observations
The data includes observations of over 2.5 million women and their market level exposure to HC (in the form of HC sales per woman) from 1970:Q1-1974:Q4. Exposure to HC through 1964:Q3 is zero since HC was not yet approved for use. The treatment is defined at the geographic market level, and hence all estimates will also be clustered at the market level.
Sample size (or number of clusters) by treatment arms
2.5 million women residing in 70 markets (clusters).
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers