♦ ♦ ♦
In this note, I discuss the findings of two reports on Reducing Abortion in America published recently by Catholics in Alliance for the Common Good. These studies examine the effect of public policies and socioeconomic factors on the abortion rate to help us understand how we can effectively lower the abortion rate. The main findings of both studies suggest that socioeconomic factors are important determinants of abortion in the United States. For example, higher male employment, more generous income assistance, and lower poverty rates are all correlated with lower abortion rates. Below, I also address specific concerns raised by Professor Michael New and suggest areas for improved research.
The initial analysis of the data mixed two different measures of abortion over time; subsequently I corrected the analysis and posted the additional results online for public scrutiny. The findings using the updated data are broadly supportive of the contention that socioeconomic factors – such as employment, poverty and income assistance – affect the abortion rate. Two findings in the original analysis – suggesting that income assistance and male employment reduce the abortion rate – remain robust in the 1990s in the updated analysis. In addition, I find that poverty is correlated with higher abortion rates in the 1990s. As pointed out in the study, the 1990s saw a dramatic decline in the overall abortion rate in the United States.
The initial analysis also found that a family cap on income assistance increases the abortion rate. In the corrected analysis, this finding remains in the quasi-experimental bivariate tests, but is not robust in the multivariate analysis. As suggested by the bivariate analysis, one reason for this discrepancy may be that I use abortion by state of occurrence and not abortion by residence. Second, in the analysis thus far, I have used a very blunt measure of state family caps that does not account for how they were implemented in different states. Future research will need to address these issues. A second finding in the initial analysis suggested that WIC reduced abortions. At the time, we were surprised by this finding, and cautioned that it needed further analysis. As I point out in the updated study, this finding was an artifact of the data we were using and did hold up in the second round of analysis. In this analysis, I only collected family cap and WIC data for the 1990s. As we collect more data, our results may yield different findings.
Finally, while the findings suggest that state laws restricting abortion have little effect on the overall abortion rate, this does not diminish Catholics in Alliance’s support for such legislation – a point that was explicitly stated in both reports. Given the findings for socioeconomic factors, though, this analysis suggests that we may be able to effectively reduce the abortion rate by addressing socioeconomic factors.
Social science research is a process that builds upon past research and updates findings as new and better data become available. In this regard, I understand that social science findings are accepted and rejected as new studies offer additional evidence. As research on abortion reduction adds data from the 2000s – when the abortion rate in the United States largely stagnated – new insights may arise and old findings may be cast in a new light. To this end, I publicly posted (in November 2008) a working paper that reports the most recent analysis of the data.
The fact remains that the abortion rate declined by over 25% during the 1990s when the economy experienced strong economic growth and gains in employment. The goal of this research is to investigate socioeconomic factors such as employment, poverty, and social supports, alongside public policies traditionally advocated by the pro-life community. As the United States Conference of Catholic Bishops (USCCB) has long noted, a comprehensive strategy to reduce abortion in the United States cannot entail a single-minded approach. The debate should not be about whether pro-life advocates must choose between advocating for restricted access to abortion or for improved social supports as strategies for reducing abortion. Given that the political climate is now much more amenable to improving social supports, though, the most effective strategy at present may be to concentrate efforts on passing legislation that improves responsible social supports for pregnant women and working families.
The evidence in both reports demonstrates that abortion rates are responsive to socioeconomic factors. Several other studies also reflect this finding, including Jones, Darroch, and Henshaw’s study (which was cited in the Catholics in Alliance report). Given the preponderance of evidence linking socioeconomic factors with the abortion rate, all of us as pro-life advocates must move beyond partisan politics and work toward a comprehensive approach to ending abortion.
The pro-life community can be unified in focusing its energies on effective strategies to reduce abortion. The Pregnant Women Support Act (PWSA) is current legislation that, as noted by the USCCB, is "truly a common-ground initiative to reduce the number of abortions in the United States." It contains important provisions to support pregnant women and parents seeking education, populations that are particularly at risk for considering abortion. In the spirit of moral consistency, pro-life advocates of all political stripes can and should back this important pro-life legislation.
Discussion of the data and empirical strategies (A detailed response to concerns raised by Michael New.)
The fact that the results for AFDC/TANF spending find different effects in the 1980s and the 1990s suggests that the qualitative differences in the implementation of income assistance programs in different time periods may be relevant. I use this finding as a starting point to help us understand how welfare reform in the 1990s may have changed the relationship between income assistance and abortion. This should not be surprising because TANF had real variation among different states in 1990s while AFDC was a national program that states were locked into during the 1980s. Further analysis of data through the early 2000s will help us understand whether the apparent abortion-reducing effect of income assistance is confined to the decade (the 1990s) of strong employment growth. I have also conducted preliminary tests to see if income assistance is more (or less) effective in low income states and in states with higher marriage rates.
The analysis of laws restricting abortion (such as parental consent laws, informed consent laws, and restrictions on Medicaid funding for abortions, partial birth abortion restrictions) finds that none of these are robustly correlated with fewer abortions. There may be a number of reasons these findings differ from New’s research:
- New uses a statistical technique that does not account for the dynamic nature of the data. Alternatively, we use an estimation technique that allows for the possibility that factors such as male employment, AFDC/TANF spending, or Medicaid funding can have both a short-term and a long-term effect on the abortion rate. In the model that New uses, the effects of the explanatory variable are assumed to occur only in the current year. This difference is important because some factors that affect the abortion rate may do so over the long term. I encourage New to use an estimation technique that accurately accounts for the dynamic structure of the data.
- New weights the data by population, which can be problematic. For example, weighting by population assumes that the measure of the abortion rate in California is ten times as accurate as the measure of the abortion rate in Mississippi or Iowa, since the population of California is over ten times that of these other two states. The results of our most current work on this project remain if: (1) we use no weights; (2) we weight the data using standard methods to address heteroskedasticity; and (3) if we weight by the log of population or the square root of the female population ages 15-44. While we do not know whether New’s results remain if he makes different assumptions about weighting the data, we have attempted to account for different possible weighting techniques. Again, I encourage New to report results with both weighted and un-weighted data. I should also note that I have been unable to replicate New’s positive findings for state restrictions on abortion access using different weights, suggesting that these results may be due entirely to his choice of population weights.
- New’s analysis leaves out states that only collect abortion data from hospitals. New intends to critique our analysis for sample selection, when in fact we have shown that our results are robust to at least three types of samples: with Kansas, without Kansas, and without states that New claims may have biased data. Thus, I encourage New to report results that are robust to various samples instead of choosing a sample that fits his results.
New claims that our results indicate that informed consent laws reduce the abortion rate. The coding of these variables means that the correct method to interpret the substantive effect of passing and enforcing an informed consent law is to add the coefficients for pass and enforce together. In our analysis, the resulting coefficient is never statistically different from zero, indicating that parental consent laws have no overall affect on the overall abortion rate. That said, the intriguing finding (for parental consent and informed consent) is that the act of passing the law appears to be correlated with higher abortions rates, while enforcing the law appears to be correlated with lower abortion rates. These results, though, hinge on only a handful of states that passed informed consent laws but did not enforce them. Adding more data from the 2000s may help shed light on this question.
Further, New claims that there are numerous peer-reviewed studies which show that state-level restrictions reduce abortion. However, in the peer-reviewed literature, there is little consensus on this issue. For example, Medhoff (2002) finds that an index of the number of state abortion restrictions has no effect on the abortion rate; and Haas-Wilson (1997) reports a similar finding. Blank et al. (1996) find evidence that restrictions on Medicaid funding for abortions is effective in reducing abortion, but that parental consent laws are not. Haas-Wilson (1993), though, finds that parental notification and informed consent can help reduce teenage abortions. The findings in these studies may differ because they analyze data from different time periods, look at different abortion data, and use different statistical techniques. An important part of the research process is to understand the underlying cause of these conflicting findings. Claims suggesting that the peer-review research has settled the debate on the effectiveness of abortion restrictions are misleading.
 Jones RK, Darroch JE and Henshaw SK (2002). "Patterns in the socioeconomic characteristics of women obtaining abortions in 2000-2001." Perspectives on Sexual and Reproductive Health.
 Professor Michael Bailey, Department of Government at Georgetown University, co-authored the first report. He requested his name be left off the updated report because he did not have time to respond to media inquiries."
 Michael New is an Adjunct Scholar at the Cato Institute, a fellow at the Witherspoon Institute, and a contributor at the Family Research Council, the Heritage Foundation, and Public Discourse. He is also an assistant professor of political science at the University of Alabama.
 Abortion data by state of residence is only collected by the CDC from 1997 onwards.
 See footnote 1.
 In the first run analysis of the data, I also found different effects in each decade, but the large abortion-reducing effect of income assistance in the 1990s swamped the opposite effect from the 1980s. In the corrected data, I still find these differential effects.
Joseph Wright is an Assistant Professor of Political Science at Penn State University, and is currently a Visiting Fellow at the Kellogg Institute, University of Notre Dame.