Politique d'aide des États-Unis d'Amérique et IVG en Afrique subsaharienne
La política de asistencia de Estados Unidos y el aborto inducido en el África Subsahariana
Eran BendavidI,*; Patrick AvilaII; Grant MillerI
IDepartment of Medicine, Stanford University, 251 Campus Drive, Stanford, California, CA, 94305, United States of America (USA)
IIStanford University School of Medicine, Stanford, USA
OBJECTIVE: To determine whether the Mexico City Policy, a United States government policy that prohibits funding to nongovernmental organizations performing or promoting abortion, was associated with the induced abortion rate in sub-Saharan Africa.
METHODS: Women in 20 African countries who had induced abortions between 1994 and 2008 were identified in Demographic and Health Surveys. A country's exposure to the Mexico City Policy was considered high (or low) if its per capita assistance from the United States for family planning and reproductive health was above (or below) the median among study countries before the policy's reinstatement in 2001. Using logistic regression and a difference-in-difference design, the authors estimated the differential change in the odds of having an induced abortion among women in high exposure countries relative to low exposure countries when the policy was reinstated.
FINDINGS: The study included 261 116 women aged 15 to 44 years. A comparison of 19942000 with 20012008 revealed an adjusted odds ratio for induced abortion of 2.55 for high-exposure countries versus low-exposure countries under the policy (95% confidence interval, CI: 1.763.71). There was a relative decline in the use of modern contraceptives in the high-exposure countries over the same time period.
CONCLUSION: The induced abortion rate in sub-Saharan Africa rose in high-exposure countries relative to low-exposure countries when the Mexico City Policy was reintroduced. Reduced financial support for family planning may have led women to substitute abortion for contraception. Regardless of one's views about abortion, the findings may have important implications for public policies governing abortion.
OBJECTIF: Déterminer si la politique de Mexico City, une politique du gouvernement des États-Unis d'Amérique interdisant le financement d'organisations non gouvernementales pratiquant ou encourageant l'avortement, était associée au taux d'interruptions volontaires de grossesse en Afrique subsaharienne.
MÉTHODES: Les femmes de 20 pays africains qui avaient subi des IVG entre 1994 et 2008 ont été identifiées dans des études démographiques et sanitaires. L'exposition d'un pays à la politique de Mexico City a été définie comme élevée (ou faible) si l'assistance reçue des États-Unis d'Amérique par habitant pour la planification familiale et la santé génésique était supérieure (ou inférieure) à la moyenne dans les pays étudiés avant le rétablissement de la politique en 2001. Utilisant la régression logistique et une approche différence en différence, les auteurs ont estimé le différentiel des probabilités d'IVG chez les femmes vivant dans les pays hautement exposés par rapport à celles vivant dans des pays faiblement exposés lorsque ladite politique a été réinstituée.
RÉSULTANTS: L'étude concernait 261 116 femmes âgées de 15 à 44 ans. Une comparaison de la période 19942000 avec la période 20012008 a révélé un rapport des cotes ajusté pour l'IVG de 2,55 pour les pays fortement exposés par rapport aux pays faiblement exposés à cette politique (intervalle de confiance de 95%, IC: 1,763,71). Sur la même période, une baisse relative du recours à des contraceptifs modernes a été notée dans les pays fortement exposés.
CONCLUSION: Le taux d'interruptions volontaires de grossesse en Afrique subsaharienne a augmenté dans les pays fortement exposés comparativement aux pays faiblement exposés lorsque la politique de Mexico City a été réintroduite. La réduction du soutien financier à la planification familiale peut avoir incité les femmes à remplacer l'avortement par la contraception. Quels que soient les points de vue sur l'avortement, les résultats peuvent avoir des implications importantes pour les politiques publiques en matière d'avortement.
OBJETIVO: Determinar si la Mexico City Policy, política gubernamental estadounidense que prohíbe la financiación a las organizaciones no gubernamentales que realizan o promueven el aborto, estuvo asociada a la tasa de abortos inducidos en África Subsahariana.
MÉTODOS: En Demographic and Health Surveys se identificaron a mujeres en 20 países africanos, que tuvieron abortos inducidos entre 1994 y 2008. La exposición del país a la Mexico City Policy fue considerada alta (o baja) si la asistencia per cápita recibida de los Estados Unidos para planificación familiar y salud reproductiva estuvo por encima (o debajo) de la media entre los países en estudio antes del restablecimiento de la política en 2001. Mediante regresión logística y un diseño de diferencia en diferencias, los autores calcularon el cambio diferencial en las probabilidades de tener un aborto inducido entre mujeres de países con alta exposición respecto a los países con baja exposición, con el restablecimiento de la política.
RESULTADOS: El estudio incluyó a 261116 mujeres de entre 15 y 44 años de edad. Una comparación de 1994-2000 con 2001-2008 reveló una proporción de probabilidades ajustadas de aborto inducido de 2,55 en los países con alta exposición frente a los países con baja exposición, durante la vigencia de la política (95% de intervalo de confianza, IC: 1,76-3,71). Hubo una reducción relativa en el uso de anticonceptivos modernos en los países con alta exposición en el mismo período.
CONCLUSIÓN: La tasa de abortos inducidos en África Subsahariana se incrementó en los países con alta exposición en comparación con los países con baja exposición, cuando se reintrodujo la Mexico City Policy. La menor asistencia financiera para planificación familiar puede haber llevado a las mujeres a sustituir la anticoncepción con el aborto. Independientemente de la propia opinión sobre el aborto, los hallazgos pueden tener implicaciones importantes para las políticas públicas que rigen el aborto.
Public policies governing abortion are contentious and highly partisan in the United States of America.1 New presidential administrations typically make a critical decision concerning abortion during their first week in office: whether or not to adopt the Mexico City Policy. First announced in Mexico City in 1984 by President Reagan's administration, the policy requires all nongovernmental organizations operating abroad to refrain from performing, advising on or endorsing abortion as a method of family planning if they wish to receive federal funding. To date, support for the Mexico City Policy has been strictly partisan: it was rescinded by Democratic President Bill Clinton on 22 January 1993, restored by Republican President George W Bush on 22 January 2001 and rescinded again by Democratic President Barack Obama on 23 January 2009.2
The Mexico City Policy is motivated by the conviction that taxpayer dollars should not be used to pay for abortion or abortion-related services (such as counselling, education or training).2 The net impact of such a policy on abortion rates is likely to be complex and potentially fraught with unintended consequences. For example, if the Mexico City Policy leads to reductions in support for family planning organizations, and family planning services and abortion are substitute approaches for preventing unwanted births (moral considerations notwithstanding), 5 then the policy could in principle increase the abortion rate. When the policy is in force, family planning organizations that ordinarily provide (or promote) abortion face a stark choice between receiving United States government funding and conducting abortion-related activities. In practice, several prominent family planning organizations, including the International Planned Parenthood Federation and Marie Stopes International, have chosen to forego United States federal funding under the policy.6
The primary aim of this study was to determine whether a relationship exists between the reinstatement of the Mexico City Policy and the probability that a sub-Saharan African woman will have an induced abortion. Specifically, we examined the association between a country's exposure to the Mexico City Policy and changes in its induced abortion rate when the policy was reinstated. Exposure was defined as the amount of foreign assistance provided to the country for family planning and reproductive health by the United States during years when the policy was not being applied. This approach enabled us to control for a variety of potential confounding factors, including fixed effects related to the country and the year of reporting, the women's place of residence and educational level, the use of modern contraceptives, and the receipt of funding for family planning activities from sources outside the United States. To the best of our knowledge, this is the first quantitative study of the association between the Mexico City Policy and abortion and the first to investigate the possibility of unintended consequences. Regardless of one's views on abortion, this lack of evidence is a critical impediment to the design of effective foreign policy and has implications for maternal mortality in places where abortion is unsafe.
We investigated the association between a country's exposure to the Mexico City Policy and the odds of abortion among women of reproductive age between 1994 and 2008 using the reinstatement of the Mexico City Policy in 2001 as a natural experiment.
Induced abortion data
We used longitudinal, individual data on terminated pregnancies collected by Demographic and Health Surveys (DHS) to estimate induced abortion rates. These standardized surveys, designed and implemented by ICF Macro, United States, in collaboration with in-country and international partners, are nationally representative surveys of women aged 15 to 49 years and are conducted approximately every six years in low- and middle-income countries.
We used every available survey with individual data on pregnancy outcomes: 30 surveys in 20 African countries. We examined data from sub-Saharan Africa because health programmes in the region receive substantial foreign assistance. In each survey, women retrospectively reported their pregnancies and pregnancy outcomes (i.e. live births and terminations) for each month during the 5 or 6 years preceding the DHS interview.
To distinguish induced from spontaneous abortions, we adopted an algorithm that uses information about the length of gestation at the time of termination, contraceptive use before the pregnancy, the desirability of the pregnancy, and maternal age and marital status at the time of the pregnancy.9 A termination was classified as induced if either: (i) it occurred following contraceptive failure; or (ii) the terminated pregnancy was unwanted (i.e. the pregnancy occurred after a live birth that was reported as unwanted or would have resulted in the number of children being more than desired); or (iii) the woman was aged under 25 years and was not married or in a union. In addition, the termination was not classified as induced if either: (iv) it occurred in the third trimester; or (v) the woman indicated that contraception had been discontinued to allow pregnancy; or (vi) the woman was married or in a union and had no children. The algorithm was tested using a DHS survey that compared the information used in the algorithm with direct questions about induced abortions (Appendix A, available at: http://bendavid.stanford.edu/research/publications.html). This approach differed from other methods to quantifying induced abortion because it provides individual-level estimates that can be used to make comparisons across countries and over time.1015The resulting data were used for our statistical analysis and to describe the longitudinal trend in the abortion rate. The abortion rate was expressed as the number of induced abortions among women aged 15 to 44 years per 10 000 womanyears.
Exposure to the Mexico City Policy
For each womanyear of observation we assigned a measure of exposure to the Mexico City Policy. First, we obtained data from the Creditor Reporting System of the Organisation for Economic Co-operation and Development (OECD) on the financial assistance provided by the United States to the country where the woman lived for family planning and reproductive health services.16,17 Then, we quantified each country's level of exposure to the Mexico City Policy using the mean value of financial assistance provided per capita by the United States for family planning and reproductive health between 1995 and 2000, when the policy was inactive (data for the period before 1995 were not available). We used this measure to create a dichotomous indicator variable that classified women as living in countries that receive financial assistance for family planning and reproductive health either above or below the median level. Our assumption was that women in countries that received a higher level of financial assistance from the United States for family planning and reproductive health between 1995 and 2000 were more exposed to the effects of a reinstatement of the Mexico City Policy.
We created another index of exposure to the Mexico City Policy using similar data on development assistance for family planning and reproductive health from the United States Agency for International Development (USAID). USAID is the main sponsor of nongovernmental organizations funded by the United States and oversees financial assistance directed towards family planning and reproductive health outside the country. We obtained information on USAID disbursements from 19962000 in the study countries and created an exposure intensity index similar to the index from the OECD data (termed "USAID" below). The entire analysis, repeated using USAID data, is presented as a sensitivity analysis below and in greater detail in Appendix A.
We used logistic regression and a difference-in-difference study design to estimate how the odds of having an induced abortion changed under the Mexico City Policy.18,19 Our approach analysed differential changes in the odds of abortion between women living in high and low exposure countries as the Mexico City Policy was reinstated in 2001. Our main dependent variable was an indicator denoting whether or not a woman had had an induced abortion in a given year. The key independent variable was the interaction between two dummy variables: whether or not the Mexico City Policy was active and whether or not a woman was living in a highly exposed country when the abortion took place. A positive regression coefficient would suggest that the odds that women living in a highly-exposed country had had an induced abortion after the Mexico City Policy was reinstated were higher than the odds of induced abortion both during the earlier period and among women in less exposed countries over the entire study period. We accounted for the interaction between the two dummy variables when calculating the odds ratio (OR) for exposure versus non-exposure.20 We also calculated marginal probabilities, which represent the differential increase in the probability of abortion among women living in highly exposed countries while the policy was in effect. Marginal probabilities were calculated at the means of the independent variables.
We controlled for country and year fixed effects, which account both for unobserved differences across countries that are stable over time and for changes in induced abortions over time that are common to all countries. All confidence intervals (CIs) are calculated using robust standard errors (SEs) clustered by country. This process incorporates the assumption that variation within countries over time is not independent when calculating SEs.2.
We also investigated underlying mechanisms that could explain the study findings. In particular, we used data from the United Nations World Contraceptive Use database to determine whether changes in the abortion rate with reinstatement of the Mexico City Policy could be linked to changes in modern contraceptive use.2
Adjusted analyses controlled for both the women's personal characteristics, such as age, marital status, place of residence (i.e. urban or rural) and educational level, and the characteristics of the country in which she was living, such as life expectancy,23 modern contraceptive use22 and the legality of abortion.24,25 To take into account the possibility that other donors compensated for restrictions imposed by the Mexico City Policy, we also adjusted for annual per capita donations for family planning and reproductive health services from all other OECD countries.1
In addition to our adjusted analyses, we also conducted several sensitivity tests. First, we reanalysed our data leaving out one survey at a time to see if our results were sensitive to a particular survey's inclusion. We repeated this procedure for each country and for each year as well. Second, we repeated our analyses using different measures of Mexico City Policy exposure, namely a continuous measure of United States assistance per capita for family planning and reproductive health and a comparable measure of exposure based on data obtained from USAID. Finally, we repeated the analyses with shorter recall windows that restricted the information we used from the DHS pregnancy calendars. These analyses are presented in Appendix A. All statistical analyses were performed with Stata 11.2 (Stata Corp, Texas, USA). The study was exempted from human subjects review by the Stanford Institutional Review Board.
Our study included data from 30 DHS surveys conducted in 20 countries between 1994 and 2008 in a total of 261 116 women (1.38 million womanyears). Table 1 (available at: http://www.who.int/bulletin/volumes/89/12/11-091660) shows the study countries and the years for which data were available; Table 2 provides descriptive statistics for our study sample and compares exposure countries and the study population with sub-Saharan Africa. The sample represented 49% of all women of reproductive age in sub-Saharan Africa. Table 2 compares the characteristics of our study sample with equivalent data for the population of sub-Saharan Africa as a whole23: there was no significant difference between the two in life expectancy at birth, mean country population, the proportion of the population living in an urban setting and country mean per capita gross domestic product.
The estimated annual induced abortion rate in the 20 study countries between 1994 and 2008 is illustrated in Fig. 1. The mean annual rate was 13 abortions per 10 000 womanyears and ranged from 8 per 10 000 in 1996 to 27 per 10 000 in 2007. The lowess (locally weighted scatterplot smoothing) curve indicates that the rate was stable between 1994 and 2001 and then rose steadily from 2002 to 2008. Overall, the induced abortion rate increased significantly from 10.4 per 10 000 womanyears for the period from 1994 to 2001 to 14.5 per 10 000 womanyears for the period from 2001 to 2008 (P = 0.01). Although the trend changed gradually, the timing of the rise is consistent with the reinstatement of the Mexico City Policy in early 2001.
Fig. 2 then shows trends in the abortion rate separately for countries with high and low levels of exposure to the Mexico City Policy. The figure suggests that the annual induced abortion rate remained stable between 1994 and 2008 in countries with low exposure to the policy, at about 1020 induced abortions per 10 000 womanyears, whereas it rose sharply after 2001 in countries with high exposure.
Table 3 presents our primary statistical results. The first row shows the differential increase in odds of abortion for women living in a highly-exposed country when the Mexico City Policy was reinstated. Each column shows a different model specification. Women living in highly exposed countries had 2.73 (95% CI: 1.953.82) times the odds of having an induced abortion after the policy's reinstatement than during the period from 1994 to 2000 or than women living in less exposed countries. After adjustments for place of residence, educational attainment, use of contraceptives and funding for family planning and reproductive health from sources other than the United States, the estimated OR dropped to 2.55 (95% CI: 1.763.71). Using a marginal probability calculation at the mean of the independent variables, we estimated that the probability of having an induced abortion exceeded the predicted odds of abortion in the absence of the policy by 9.9 per 10 000 woman-years (P < 0.001). A likelihood ratio test of the principal model compared with a model of women living in less exposed countries suggested a significant difference in odds between the less exposed and the highly exposed populations (P < 0.001): the OR of an abortion using USAID data for determining country exposure was 1.82 (95% CI: 1.093.02). Additional results using USAID data for country exposure are available in Appendix A.
We also examined whether the number of years during which the Mexico City Policy was in force was associated with the rate of induced abortion. We estimated that, for each additional year under the Mexico City Policy, the odds of having an induced abortion were 1.21 times higher among women living in a country with a high level of exposure between 2001 and 2008 than among women in the reference groups (OR: 1.21; 95% CI: 1.101.37).
We repeated our analysis using a continuous measure of exposure to the policy: the mean per capita assistance for family planning and reproductive health provided to each country by the United States between 1995 and 2000. We found a positive but nonsignificant association between each additional United States dollar of per capita assistance and the odds of an induced abortion (OR: 1.07: 95% CI: 0.91130).
We investigated whether our findings were influenced by the recall period, i.e. the time between the pregnancy outcome and the date of the DHS in which it was recorded. The reliability of recall may decrease with time. We found that the length of recall window used to construct our sample did affect the study's power but did not result in substantial bias (Appendix A, Fig. A).
Finally, to explore underlying behavioural responses that could explain our main results, we investigated whether exposure to the Mexico City Policy was associated with modern contraceptive use. The variation in contraceptive use between 1994 and 2008 in countries with high or low levels of exposure to the policy is shown in Fig. 3. The increase in contraceptive use proceeded at a slower pace after 2002 in countries with a high level of exposure, whereas in countries with a low exposure the increase continued at the same pace. We repeated our main regression analysis using the prevalence of modern contraceptive use as the dependent variable and found this to be 1.8% lower (95% CI: 0.13.4) under the Mexico City Policy than expected in the light of trends in low exposure countries.
Some American presidential administrations care deeply about curtailing access to abortion, whereas others seek to promote it. Regardless of one's views, an understanding of the relationship between the Mexico City Policy and abortion rates in developing countries is important for foreign policy decisions. To the best of our knowledge, this study is the first quantitative analysis of the policy's possible implications for women living in countries that depend heavily on development assistance for family planning and reproductive health services.
Our study found robust empirical patterns suggesting that the Mexico City Policy is associated with increases in abortion rates in sub-Saharan African countries. Although we are unable to draw definitive conclusions about the underlying cause of this increase, the complex interrelationships between family planning services and abortion may be involved. In particular, if women consider abortion as a way to prevent unwanted births, then policies curtailing the activities of organizations that provide modern contraceptives may inadvertently lead to an increase in the abortion rate.
Several observations strengthen this conclusion. First, the association is strong: the odds of having an abortion in highly exposed countries were more than twice the odds observed in the reference groups. Second, there is broad agreement among our aggregate graphical analysis and both unadjusted and adjusted statistical analyses, and our main findings are robust across a variety of sensitivity analyses. Third, the timing of divergence between high and low exposure countries is coincident with the policy's reinstatement: in high exposure countries, abortion rates began to rise noticeably only after the Mexico City Policy was reinstated in 2001 and the increase became more pronounced from 2002 onward. Finally, our findings are consistent with those of previous studies on the relationship between family planning activities and abortion.
Given the potential implications of our study, its limitations deserve careful attention. First, our abortion rate estimates are lower, on average, than rates reported elsewhere,26,27 most of which used a multiplier to correct for underreporting. However, if the level of underreporting is not related to the level of exposure to the Mexico City Policy or to the timing of the policy's implementation, then our approach to estimating changes associated with the policy is acceptable given our statistical model, since we were interested in changes over time rather than the absolute abortion rate. Second, our results may depend on the specific countries included in the study and may not be generalizable to other countries receiving financial assistance from the United States. Nonetheless, our results were not affected by the omission of any single DHS or country, nor were they influenced by using an alternative data source to define countries with a high or low level of exposure to the Mexico City Policy (Appendix A). Third, changes in presidential administration often bring changes in foreign relations, and a country that received a high level of aid from the United States for family planning before 2001 may have become a less-favoured nation and received less aid under the subsequent administration. Although we cannot rule out this possibility, this mechanism would support our premise that funding for family planning services has a paradoxical effect on abortion outcomes. Fourth, the observed rise in abortion rates may be a result of an ideational change in favour of smaller families that took place in countries with a high level of exposure to the Mexico City Policy when it was reinstated.28,29 However, this change is not likely to have occurred over such a short period of time. Finally, the pregnancy calendars we used to estimate abortion rates are a relatively new component of the DHSs and further research is needed to fully establish their reliability. Continued monitoring of the trend in abortion rates since the policy was last rescinded will provide more data on its impact.
We conclude by noting that beyond the scope of this study, the relationship between the Mexico City Policy and reproductive health deserves attention. With growing international emphasis on reducing maternal mortality, in keeping with Millennium Development Goal 5, our findings suggest that this United States policy may have unrecognized and unintended health consequences.
The authors thank Paul Blumenthal at Stanford University, Kristin Wilson at Harvard Business School and Manisha Shah at the University of California in Irvine.
Funding: Eran Bendavid is supported by the National Institute of Allergy and Infectious Diseases (K01-AI084582) and by the George Rosenkranz Prize for Health Care Research in Developing Countries. Grant Miller receives support from the National Institute of Child Health and Development and the National Institute of Aging. Both receive support from the Center on the Demography and Economics of Health and Aging (P30-AG17253). The funding agencies played no part in the study or its analysis and reporting.
Competing interests: None declared.
1. Annas GJ. Abortion politics and health insurance reform. N Engl J Med 2009;361:258991. doi:10.1056/NEJMp0911513 PMID:19955513
2. United States Agency for International Development [Internet]. Restoration of the Mexico City Policy concerning family planning. Washington: USAID; 2011. Available from: http://www.usaid.gov/bush_pro_new.html [accessed 14 Sept 2011] .
3. The White House [Internet]. Mexico City Policy voluntary population planning: memorandum for the Secretary of State and the Administrator of the United States Agency for International Development. Washington: The White House; 2009. Available from: http://www.whitehouse.gov/the_press_office/MexicoCityPolicy-VoluntaryPopulationPlanning/ [accessed 31 August 2011] .
4. AID Family Planning Grants/Mexico City Policy. Washington: The White House; 1993. Available from: http://clinton6.nara.gov/1993/01/1993-01-22-aid-family-planning-grants-mexico-city-policy.html [accessed 31 August 2011] .
5. Bongaarts J, Westoff CF. The potential role of contraception in reducing abortion. Stud Fam Plann 2000;31:193202. doi:10.1111/j.1728-4465.2000.00193.x PMID:11020931
6. Boseley S. Condom supply to Africa hit by US abortion policy. The Guardian. 2003 25 September. Available from: http://www.guardian.co.uk/world/2003/sep/25/aids.usa [accessed 22 August 2011] .
7. Crossette B. Implacable force for family planning. The New York Times. 2002 30 July. Available from: http://www.nytimes.com/2002/07/30/health/implacable-force-for-family-planning.html [accessed 22 August 2011] .
8. AIDS and abortion policy. Pregnant pause: how one policy undermines another. The Economist. 2003 25 September. Available from: http://www.economist.com/node/2091350 [accessed 22 August 2011] .
9. Magnani RJ, Rutenberg N, McCann H. Detecting induced abortions from reports of pregnancy terminations in DHS calendar data. Stud Fam Plann 1996;27:3643. doi:10.2307/2138076 PMID:8677522
10. Henshaw S, Singh S, Oye-Adeniran B, Adewole I, Iwere N, Cuca Y. The incidence of induced abortion in Nigeria. Int Fam Plan Perspect 1998;24:15664. doi:10.2307/2991973
11. Huntington D, Mensch B, Toubia N. A new approach to eliciting information about induced abortion. Stud Fam Plann 1993;24:1204. doi:10.2307/2939205 PMID:8390117
12. Houzard S, Bajos N, Warszwawski J, de Guibert-Lantoine C, Kaminski M, Leridon H et al. Analysis of the underestimation of induced abortions in a survey of the general population in France. Eur J Contracept Reprod Health Care 2000;5:5260. doi:10.1080/13625180008500370 PMID:10836663
13. Ahmed M, Rahman M, Van Ginneken J. Induced abortion in Matlab, Bangladesh: trends and determinants. Int Fam Plan Perspect 1998;24:12832. doi:10.2307/3038209
14. Stover J. Revising the proximate determinants of fertility framework: What have we learned in the past 20 years? Stud Fam Plann 1998;29:25567. doi:10.2307/172272 PMID:9789319
15. Rossier C. Estimating induced abortion rates: a review. Stud Fam Plann 2003;34:87102. doi:10.1111/j.1728-4465.2003.00087.x PMID:12889341
16. OECD.StatExtracts [Internet]. Creditor Reporting System, full. Paris: Organisation for Economic Co-Operation and Development; 2010. Available from: http://stats.oecd.org/Index.aspx?DatasetCode=CRSNEW [accessed 23 September 2011] .
17. Ravishankar N, Gubbins P, Cooley RJ, Leach-Kemon K, Michaud CM, Jamison DT et al. Financing of global health: tracking development assistance for health from 1990 to 2007. Lancet 2009;373:211324. doi:10.1016/S0140-6736(09)60881-3 PMID:19541038
18. Shadish W, Cook T, Campbell D. Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton Mifflin Company; 2002.
19. Meyer BD. Natural and quasi-experiments in economics. J Bus Econ Stat 1995;13:15161. doi:10.2307/1392369
20. Ai C, Norton EC. Interaction terms in logit and probit models. Econ Lett 2003;80:1239. doi:10.1016/S0165-1765(03)00032-6
21. Bertrand M, Duflo E, Mullainthan S. How much should we trust differences-in-differences estimates? Q J Econ 2004;119:24975. doi:10.1162/003355304772839588
22. World contraceptive use 2009. New York: United Nations Population Division, Department of Economic and Social Affairs; 2009.
23. World Development Indicators Online (WDI). Washington: The World Bank; 2011. Available from: databank.worldbank.org/ [accessed 14 Sept 2011] .
24. Bloom D, Canning D, Fink G, Finlay J. Fertility, female labor force participation, and the demographic dividend. J Econ Growth 2009;14:79101. doi:10.1007/s10887-009-9039-9
25. Abortion policies: a global review. New York: United Nations, Population Division; 2002. Available from: http://www.un.org/esa/population/publications/abortion/index.htm [accessed 31 August 2011] .
26. Henshaw SK, Singh S, Haas T. The incidence of abortion worldwide. Int Fam Plann Persp 1999;25:S308. doi:10.2307/2991869 PMID:14627053
27. Sedgh G, Henshaw S, Singh S, Åhman E, Shah I. Induced abortion: estimated rates and trends worldwide. Lancet 2007;370:133845. doi:10.1016/S0140-6736(07)61575-X PMID:17933648
28. Henshaw SK, Singh S, Haas T. Recent trends in abortion rates worldwide. Int Fam Plan Perspect 1999;25:448. doi:10.2307/2991902
29. Rahman M, DaVanzo J, Razzaque A. Do better family planning services reduce abortion in Bangladesh? Lancet 2001;358:10516. doi:10.1016/S0140-6736(01)06182-7 PMID:11589936
Submitted: 7 June 2011
Revised version received: 11 August 2011
Accepted: 16 August 2011
Published online: 27 September 2011