(Note: an opinion piece summarizing these conclusions is available here.)

Contents

• Summary
• Background
• Review of Stassen's quantitative analysis regarding post-2000 abortion trends
• Subsequent AGI and CDC analyses
• Stassen's thesis regarding socio/economic policy and abortion
• Stassen's statements regarding his pro-life position
• Conclusions
• Sources

Summary

Shortly before the November 2004 U.S. presidential election, Glen Stassen published opinion pieces claiming that Bush administration policies had produced an increase in U.S. abortions since 2000. These claims garnered some media attention, despite the fact that other analysts soon identified flaws in the claims. I evaluate Stassen's claims here, for two reasons: they have been cited by leading politicians in public discussion, and Stassen cited data from my website in reference to his claims.

General conclusions are as follows:

• Stassen claimed that his analysis of then-available data showed that U.S. abortions had increased from 2000 to 2001, from 2001 to 2002, and from 2002 to 2003. His analysis is flawed in both data collection and in statistical approach, negating any positive claims of a detected increase in abortions. Specifics include a demonstrably biased sample of states, failure to include available data for 10 states, and use of data from two states with known reporting problems. More complete studies by the AGI and CDC have since been published, showing a continued decrease in annual abortions.
• Stassen claimed that this purported abortion increase was the result of Bush administration economic and social policies. He provides little data to support this claim; what supportive data he does provide is sometimes inaccurate. Actual data is generally inconsistent with his thesis, and the logical case for his thesis is flawed. In particular, government data refutes his alleged correlations between abortion rates and the following: unemployment rates, marriage rates, minimum wage, low income, health insurance coverage, and education spending. Stassen's additional claims regarding 1990s abortion trends, reasons given by women for abortions, infant mortality, and Bush administration economic policies contain inaccuracies.
• In follow-up articles, Stassen has defended his approach and his overall thesis, although he has partially acknowledged errors in his quantitative conclusions. For the most part, these subsequent articles do not acknowledge the principle defects of his analysis.
• Stassen describes himself as "consistently pro-life" and denies signing a document endorsing the Roe v. Wade decision. The document in question, the 1977 "Call to Concern", does state "We support the Supreme Court decisions of 1973 which had the effect of removing abortion from the criminal law codes" and was signed by Stassen.

Background

Dr. Glen H. Stassen is a professor of Christian ethics at Fuller Theological Seminary in Pasadena, California. In October 2004 he published an article in Sojourners magazine describing his analysis of limited post-2000 state-level abortion data. He concluded, based on analysis of data from 16 states, that annual numbers of U.S. abortions began rising in 2001-2002 after two decades of a downward trend. Further, he claimed that this was the result of Bush administration social and economic policies, contrasting the stated administration objectives of reducing abortions.

Various observers (particularly those on the political right) quickly criticized Stassen's conclusions, many pointing out the unreliability of the approach used in his quantitative analysis. This included, notably, researchers at the National Right to Life Committee. Stassen released material defending his conclusions, in particular releasing the state-level data he used in his analysis. In later articles, Stassen has expanded discussion of his thesis connecting social/economic policies and abortion rates.

A number of individuals on the political left cited Stassen's articles in criticizing the Bush administration, both before and after the presidential election. For example, on 24 January 2005, Senator Hillary Rodham Clinton repeated Stassen's data in a critique of the Bush administration. The notion that abortions have increased has taken on a life of its own: on 22 May 2005, Howard Dean claimed that abortions had increased 25% under Bush--a completely erroneous claim, and not one made by Stassen.

In response to the debate produced by the claimed abortion increase, the Alan Guttmacher Institute released a preliminary analysis of data for 2001 and 2002, concluding that abortions had decreased each year from the year before. The CDC had previously released data for 2001, showing a decrease in abortions from 2000.

Stassen has discussed his analysis and/or thesis in multiple forums since Oct. 2004. They are listed below with codes I will use in referencing them.

[S1] 13 Oct. 2004, "Pro-Life? Look at the Fruits", SojoMail, and Oct. 2004, Sojourners, on line at Fuller Theological Seminary [http://www.fullerseminary.net/sot/faculty/stassen/Resource_files/Prolife4.htm]. [SK2] with Gary Krane, 17 Oct. 2004, "Why abortion rate is up in Bush years," Houston Chronicle, on line at FactCheck.org [http://www.factcheck.org/UploadedFiles/HoustonChronicle.com-Why-abortion-rate-is-up-in-Bush-years.pdf]. [S3] 22 Oct. 2004, untitled, on line at Between Two Worlds [http://theologica.blogspot.com/2004/10/prof-stassen-responds.html]. [S4] 27 Oct. 2004, "Sources and calculations for the increase in abortion rates in my prolife op/ed," on line at Fuller Theological Seminary [http://www.fullerseminary.net/sot/faculty/stassen/Resource_files/Vital%20Statistics%20At%20a%20Gla.htm]. [S5] 31 Jan. 2005, "Professor Glen Stassen Interview," Malkin(s)Watch, on line [http://malkin-watch.blogspot.com/2005/01/professor-glen-stassen-interview.html]. [S6] 22 Feb. 2005, "Supporting parents," Christian Century, on line at Find Articles [http://wwww.findarticles.com/]. [S7] 25 May 2005, untitled, on line at FactCheck.org [http://www.factcheck.org/UploadedFiles/Stassen-Response.pdf]. [S8] 15 June 2005, "Reaffirming My Op Ed Article that Prolife Needs to Pay Attention to Economic Justice," and 29 June 2005, "The pro-life movement and economic justice," Sojomail, on line at Fuller Theological Seminary [http://www.fullerseminary.net/sot/faculty/stassen/Resource_files/Agree%20on%20Abortions.pdf].

(Of the articles listed above, two cite my website: S4 on 27 Oct. 2004 and S8 on 15 June 2005.)

Review of Stassen's quantitative analysis regarding post-2000 abortion trends

In the United States, abortion statistics are not directly compiled at the national level. Individual states implement their own practices regarding data collection. These vary in terms of completeness, detail, and timeliness. Currently, four states do not collect such statistics, and several others do not publish them routinely. The Centers for Disease Control compiles data reported to it by those states that gather statistics; these reports are generally issued annually. This data is incomplete, since abortion providers are generally not compelled to complete reporting to their respective states. The Alan Guttmacher Institute, a research group advocating access to abortion, compiles its own statistics using surveys of abortion providers. These figures are generally acknowledged (by both supporters and opponents of abortion) as the most authoritative nationwide data. The AGI reports are published periodically, every 2-3 years in recent years. From 1988 to 1997, CDC figures averaged 11.3% lower than AGI figures. Since 1997, CDC figures are lower because data from 3-4 states are not available.

The latest AGI analysis covered the year 2000 and was published in early 2003. In November 2003 the CDC released its analysis for the year 2000. (CDC data for 2001 was released in November 2004, after Stassen's first articles.)

There are a number of problems with Stassen's approach; I will treat these in turn below.

1. Did Stassen use accurate and appropriate data? In S4, Stassen reports abortion figures for the 16 states of his sample. This data is reproduced below:

   state	2001	2002	2003	net change, 2001-2002	type of data
   Alabama	11,852	10,648		-1,204	residents, in and out of state?
   Arizona	8,226	10,397		+2,171	residents, in state
   Colorado	4,463	7,757	9,852	+3,294	(residents, in state?)
   Florida	85,589	87,964		+2,375	all occurrences
   Idaho	728	829		+101	all occurrences
   Illinois	46,546	46,945		+399	all occurrences
   Kentucky	3,502		3,621	+60	(residents, in state?)
   Michigan	28,220	29,231	29,540	+1,011	all occurrences
   Minnesota	14,833	14,186	14,024	-647	all occurrences
   Missouri	7,797	8201		+404	all occurrences
   Ohio	37,464	35,830		-1,634	all occurrences
   Pennsylvania	36,820	35,167		-1,653	all occurrences
   South Dakota	759	685		-74	residents, in state
   Texas	77,537	79,929		+2,392	all occurrences
   Washington	25,998	25,446		-552	residents, in and out of state
   Wisconsin	10,925	10,489	10,557	-436	all occurrences

   The total increase in abortions from 2001 to 2002 from this data is 6,007. In S1 and S2, Stassen reported this increase as 7,869. In S3, his first response to the NRLC article, he acknowledged an error in reporting data for South Dakota and Wisconsin (apparently the amount of decrease for these states had been counted as an increase), and gave a corrected total increase of 6,849. In S4 he gave the data above and gave a total increase of 6,207--still incorrect, even based on this data. In S5 and S6, he again revised his total increase downward to 5,855. In S8, he reports the increase as "about 6,000." Stassen has revised his original total downward three times now.

   Originally, Stassen claimed that abortions had increased in 12 states and decreased in 4 states from 2001 to 2002. After correcting his errors with South Dakota and Wisconsin, this became 10 states with increases and 6 states with decreases.

   Importantly, Stassen does not consistently use the same type of data. For 10 states he uses data by state of occurrence (abortions performed in state, both to residents and nonresidents). For 4 states he uses data by state of residence--for 2 states abortions performed in state to residents only, and for 2 states abortions performed both in and out of state to residents only. For two states (Colorado and Kentucky), the data used by Stassen has not been published; this data may be by state of residence, but this is not known for certain.

   In the United States, many women cross state lines to obtain abortions. According to CDC data for 2001, about 8.7% of U.S. abortions are obtained outside the state of residence. Residency of women obtaining abortions is not consistently reported, introducing issues with the use of residency-based data. Data by state of occurrence, however, can fluctuate more due to availability or restrictions on abortions in neighboring states.

   Depending on whether data by state of occurrence or by state of residence is used, one can obtain a total net change for the Stassen sample anywhere from 5,155 to 7,595. If occurrence data is preferred, the net change is 6,357; if residence data is preferred, the net change is 7,168 (this assumes Stassen's figures for Colorado and Kentucky are by state of residence); if residence data including out of state abortions are preferred, the net change is 6,314. The table below gives the various available data for these 16 states.

   state	all occurrences		residents, in state		residents, in/out of state
   	2001	2002	2001	2002	2001	2002
   Alabama	13,382	12,249			11,852	10,648
   Arizona	8,302	10,677	8,226	10,397	8,601
   Colorado	4,633		4,463	7,757	3,977
   Florida	85,589	87,964			431
   Idaho	738	829	712	800	1,473	1,493
   Illinois	46,546	46,945	41,523	42,655	42,427
   Kentucky	3,764		3,502	3,562	6266
   Michigan	28,220	29,231	27,208	28,229	27,534
   Minnesota	14,833	14,186	13,448	12,901	13,971
   Missouri	7,797	8,201			12,266	12,250
   Ohio	37,464	35,830	34,174	32,687	35,059
   Pennsylvania	36,820	35,167	35025	33,503	37,498
   South Dakota	895	826	759	685	1,045
   Texas	77,537	79,929	74,101	76,278	74,237	76,278
   Washington	25,620	25,148	24,353	23,960	25,998	25,446
   Wisconsin	10,925	10,489	10,703	10,305	12,279

2. Are figures from Arizona and Colorado reliable?

   Most of Stassen's total net change is from Arizona and Colorado; these two states give a net change (from Stassen's choice of figures) of 5,465. In his data, these two states show increases from 2001 to 2002 of 26.4% and 73.8%. Arizona's figure for 2001 is also 29.5% below that of 1999 and Colorado's is 86.6% below that of 1997. Both states, then, exhibit very dramatic drops in official figures for a few years, then a return to previous values. The increase Stassen obtains results from comparing this temporary minimum in 2001 to subsequent more typical values.

   Given the previously noted problems with completeness, someone familiar with this data should suspect underreporting in the 2001 figures. This is confirmed by the respective states, as noted in the first NRLC review. The Arizona Department of Health Services notes, of the 2002 increase, "It is unclear whether this increase in the number of reported abortions represents a true increase in the actual number of abortions performed, or, perhaps, a better response rate of providers of non-surgical (so called medical) terminations of pregnancy." The Colorado report is even more specific, noting that underreporting by abortion providers--in violation of regulatory requirements--was a problem in 2001, such that regarding the increase in 2002 "No one could or should conclude that this anticipated increase in the rate of reported terminations reflects an increase in the true rate."

   Thus, most of the increase found by Stassen is the result of data from two states, data known to be flawed. Eliminating these two states changes Stassen's increase to only 390, about 5% of what he originally claimed.

   In response, Stassen has claimed that he is taking a scientific approach by not rejecting any values, keeping to a uniform approach. While such an approach might be uniform, it is unscientific: incorporating flawed measurements in an analysis, particularly ones known to be flawed, can only produce unreliable results. Even without the reports from the states warning that the 2001-2002 increase is not real, the extraordinarily high sensitivity of Stassen's results to only two states undermines the ability to extrapolate such a small sample to the entire nation. This is also demonstrated below.

3. Are these states representative of the United States? Stassen's approach assumes that the sample of 16 states is representative. Stassen claims in S4 that, if anything, this sample would give a lower estimate: "The Midwest being over-represented also over-represents states whose number of abortions decreased... Thus if anything the national increases are likely to be greater than I extrapolated."

   However, if one compares cumulative abortion trends in the 1990s for these 16 states to the U.S. as a whole, one finds that the sample is biased against a downward trend. Using AGI data for 1991, 1995, and 2000, below are the trends for the Stassen sample and for the entire U.S.:

   sample	1991-1995	1995-2000	1991-2000
   Stassen sample	-10.16 %	-0.95 %	-11.02 %
   United States	-12.66 %	-3.41 %	-15.64 %

   If the data is annualized, to give average annual percentage decrease in each time period, these results are:

   sample	1991-1995	1995-2000	1991-2000
   Stassen sample	-2.64 %	-0.19 %	-1.29 %
   United States	-3.33 %	-0.69 %	-1.87 %

   For the preceding decade, the rate of decline in abortions has been significantly greater in the entire U.S. than in the Stassen sample. This data, one simple indicator of the validity of Stassen's sample, was readily available by 2003.

4. Does Stassen make use of all available data? In S4, where Stassen defends the inclusion of Arizona and Colorado, he also says he used all available data and reports my website as a source. In May 2004 my website included 2001-2002 abortion data for ten states not included in Stassen's sample. These are Arkansas, Georgia, Iowa, Kansas, Montana, North Carolina, Oregon, Utah, Vermont, and Virginia. Of these ten states, 6 posted decreases and 4 posted increases in abortions. Data for these states (as available in 2004) is as follows:

   state	all occurrences		residents, in state		residents, in/out of state
   	2001	2002	2001	2002	2001	2002
   Vermont	1,513	1,635	1,297	1,411
   Montana	2,350	2,248	2,063	2,034
   Utah	3,594	3,524	3,372	3,300
   Arkansas	5,924	5,316	4,981	4,511
   Iowa	5,720	6,230	5,093	5,384
   Kansas	12,422	11,844	6,419	6,298
   Oregon	14,272	13,172	12,469	11,536
   Virginia					23,968	24,347
   Georgia			30,064	30,432
   North Carolina		29,229	27,096	25,883

   The total net change for these 10 states is a decrease of 2,292 if data by state of occurrence is preferred, or a decrease of 1,686 if data by state of residence is preferred. By failing to include these states for which data was available in 2004, Stassen introduced a bias in favor of increasing abortions.

   If these 10 states are included with Stassen's sample, the total net change is an increase of 4,065 (preferring occurrence data) or 5,482 (preferring residence data). These figures are of course mostly influenced by the flawed Arizona and Colorado figures. Eliminating these two states, the total net change is a decrease of 1,400 abortions (preferring occurrence data) or an increase of 17 (preferring residence data). The sample available in 2004, excluding Arizona and Colorado, comprised 24 states of which 12 showed increases and 12 showed decreases.

5. Does Stassen appropriately extrapolate his sample to a national total?

   In S4, Stassen indicates he scaled his derived total net change by a factor of 3.265 to obtain the estimated nationwide net change. This ratio appears to be the ratio of the last reported AGI total for the U.S. to the total of his state-reported data. Such scaling assumes no change in the level of abortion reporting from year to year.

6. Is Stassen's estimate as robust as he claims?

   In S4, Stassen states, "A Z-test of statistical significance of the sixteen states for the one year that I reported, representing about thirty million women of child-bearing age, suggests greater than 99.9999% confidence that they represent the fifty states." This is a serious misapplication of this statistical test. Stassen's results are not actual measurements involving 30 million women, they are limited sampling involving about 400,000 reported abortions. Further, what he is actually working with is a set of measurements known to be incomplete. The z-test method involves certain assumptions regarding the data, assumptions which may not be valid. As discussed above, data was available in 2003 which demonstrated that the Stassen sample does not represent the 50 states.

   To illustrate, consider the reported percentage change for the original 16 states from 2001 to 2002 to be our measurements. In this case, there is a 42% probability of a decrease instead of an increase, and a 53% chance that Stassen has overestimated the change. (This is not claimed to be the appropriate statistical treatment, but merely an illustration of the radical difference between one possible approach and Stassen's incorrect approach.)

7. How do Stassen's results compare to results obtained from a larger sample of states?

   Finally, here are results of applying Stassen's method to currently available data. This is done to illustrate the high dependence of results on what states are included in the sample. Given first are derived changes in abortion numbers from 2000 to 2001:

   sample		total net change, 2001-2002
   description	# of states	sample states	extrapolated for U.S.
   all available now	48	-7,827	-9,584
   only Stassen 16	16	-1,442	-3,834

   The data is mostly complete, including 47 states and the District of Columbia. State-level data for Alaska, California, and New Hampshire is not available. The prediction from the full sample compares well to the estimate from AGI, discussed in the next section, despite the lack of data from California (which accounted for 18% of U.S. abortions in 2000, according to AGI). Note that a prediction based only on Stassen's sample of 16 states significantly underestimates the decrease. This shows again that Stassen's sample is not representative of the U.S. as a whole.

   Here are derived changes in abortion numbers from 2001 to 2002:

   sample		total net change, 2001-2002
   description	# of states	sample states	extrapolated for U.S.
   all available now	38	1,785	2,458
   all available less AZ/CO/WY	35	-3,239	-4,622
   Stassen 16	16	6,612	17,579
   Stassen 16 less AZ/CO	14	1,113	3,174
   all available in 2004 (Stassen 16 +10)	26	4,368	8,949
   all available in 2004 less AZ/CO	24	-1,131	-2,445

   Note that the results are dramatically affected by the inclusion of states with reporting problems: Arizona, Colorado, and Wyoming (for Wyoming, only some out-of-state data is available for 2002). Also, the larger samples predict a smaller change than Stassen's sample.

   The next table gives derived changes in abortion numbers from 2002 to 2003:

   sample		total net change, 2002-2003
   description	# of states	sample states	extrapolated for U.S.
   all available now	34	2,209	3,779
   all available less CO	33	114	199
   Stassen 5	5	2,367	33,649
   Stassen 5 less CO	4	272	4,644

   Here, there is a dramatic effect from the inclusion of Colorado, where data still implies reporting deficiencies. Also, the full sample available today gives a very different prediction than that made from Stassen's sample of five states.

   The next table gives derived changes in abortion numbers from 2003 to 2004:

   sample		total net change, 2003-2004
   description	# of states	sample states	extrapolated for U.S.
   all available now	7	-5,366	-100,022
   all available now less OR	6	-2,431	-59,740

   Oregon data for 2004 is incomplete; note that the results are significantly affected by inclusion of Oregon. If Stassen's method were valid, available data predicts a very large decrease in abortions in 2003-2004. The method, of course, is not reliable for such a small sample of states. Neither an increase or a decrease can be predicted based on the limited data currently available.

Subsequent AGI and CDC analyses

The month following Stassen's first articles, the CDC released its review of abortion incidence for 2001. CDC data for 2000 and 2001 does not include Alaska, California, and New Hampshire, which did not report data. For the remaining 47 states plus the District of Columbia, the CDC reported 857,475 abortions in 2000 and 853,485 in 2001, a decrease of 3,990 abortions or 0.5%. If the same rate of decline is applied to AGI's more complete estimate of abortions in 2000 of 1,313,000, the decrease would be 6,100 abortions.

In S4, Stassen reports applying his analysis to 10 states for which he had data "handy" for 2000 and 2001, and obtained an increase of 4,067 abortions in 2001. Based on CDC results, Stassen's method does not give the correct trend for 2000-2001.

On 18 May 2005, the Alan Guttmacher Institute released an analysis by Lawrence Finer and Stanley Henshaw estimating U.S. abortion incidence in 2001 and 2002. The results were preliminary projections based on available state-reported data, since the periodic AGI provider survey had not been completed. The AGI press release indicated that these preliminary results were being provided to address the debate over whether abortions had increased since 2000.

AGI concluded that abortions dropped from 1,313,000 in 2000 to 1,303,000 in 2001, and again to 1,293,000 in 2002, a decrease of 10,000 each year. In their report, they noted that reporting in Arizona, Maryland, Nevada, and the District of Columbia is very incomplete or inconsistent. As a check, they applied the method to data from 1996 to 1997 and obtained a 1.0% error, but noted that abortions were declining at a greater rate then so "Therefore, it is likely that our current projections will have a smaller margin of error." They report average annual declines of 3.4% per year from 1992 to 1996, of 1.2% per year from 1996 to 2000, and of 0.9% per year from 2000 to 2002, "suggesting that the last two years reflect a continuation of the trend of the late 1990s, albeit at a slightly slower rate of decline."

The associated AGI press release implies that the analysis was conducted in response to the debate caused by Stassen's claims: "Yet demands for more recent data, resulting in part from media reports, opinion pieces and public speeches speculating that abortion has increased as a result of Bush administration policies, have prompted the Institute to analyze available government data as an interim measure until another provider census can be conducted." But the AGI disputes not only Stassen's quantitative conclusions, but his suggested relation between abortion rates and administration policy. Sharon Camp, AGI president and CEO, said "It takes time for political decisions to be reflected in the statistical data, so it is too soon to tell what the impact of Bush administration policies will be on U.S. abortion rates."

Stassen responded to the AGI results in S7. Stassen cites their 1% margin of error for 1996-1997 estimates and says "So the decline of 0.9% [from 2001 to 2002] is within the margin of error." He does not acknowledge that his claimed 1.5% increase is well outside this margin of error, nor that the authors believed the margin of error for 2000-2002 would be less than 1.0%.

Regarding AGI's reported decrease in the abortion rate, Stassen says the 1% margin of error is "larger than their estimated change in abortion ratios for 2002 of 0.3." But this change in abortion ratio is a decrease of 1.2% from 2001. He interprets the AGI-reported decreases as a "stall" particular to the Bush administration, even though AGI notes that this low rate of decrease is little changed since the late 1990s.

Stassen's thesis regarding socio/economic policy and abortion

Stassen's broader thesis is that government social and economic policies are driving forces in the rate of abortions. He makes over a dozen specific claims of casual relations between various socio/economic measures and abortion rates. In most cases the data argue against his stated or implied relationship. Only in a few cases does he provide data; in many of these cases, the data is not accurately described. This section will review the various claims.

1. Abortion trends in 1990s: Stassen discusses the trends in abortion numbers and rates in the 1990s, but he mostly mischaracterizes the trends.

   In S1, Stassen says abortion rates declined 17.4% in the 1990s "mostly during the latter part of the decade." The actual decline, from AGI figures, was 22.3% from 1990 to 2000 and was mostly in the first half of the decade--17.9% from 1990 to 1995 and only 5.3% from 1995 to 2000. In SK2, he calls the decline "steady" through the decade, also incorrect.

   In S3, he states "Had the average annual decrease in the years prior to Bush continued, we would have instead expected a decrease of 28,000 fewer abortions." The most recent year in which there was a decrease in abortions as high as this was 1995, when AGI estimates a decrease of 64,000 abortions over the previous year. From AGI figures, the average annual decrease in abortion incidence was 49,900 from 1991 to 1995 and 9,300 from 1996 to 2000. The AGI estimated decreases in 2001 and 2002 are 10,000 each year, about the same as in the second half of the 1990s.

   In S5, Stassen cites CDC data to claim that the abortion rate was flat during the Bush administration from 1989 to 1992, then dropped during the Clinton administration. Stassen notably does not use the same AGI figures he cited previously in the same article; the AGI figures are more complete and give a slightly different story. The AGI-reported abortion rate dropped an average 1.4% per year during the Bush administration (from 26.8 in 1989 to 25.7 in 1992), dropped an average 3.6% per year during Clinton's first term (from 25.0 in 1993 to 22.4 in 1996), then dropped an average 1.3% per year during Clinton's second term (from 22.4 in 1996 to 21.3 in 2000).

   Even after confronted with the AGI/CDC conclusions that abortions continued to decline after 2000, Stassen continues to contrast this "stall" with the "dramatic" decrease in abortions in the 1990s. However, the fact is that the decrease was mostly in the first half of the 1990s, and the post-2000 rate of decrease is quite similar to the rate of decrease in the second half of the 1990s. This is shown in the graph below, of annual percentage changes in abortion rates and abortion incidence (from AGI figures).

   

2. Reasons given by women for abortions: In S1, Stassen cited Minnesota Citizens Concerned for Life (MCCL) in stating that "two thirds of women who abort say they cannot afford a child" and "half of all women who abort say they do not have a reliable mate." In SK2, the claim becomes "two-thirds of women who have abortions cite 'inability to afford a child' as their primary reason." The MCCL web site, citing an AGI page, in turn citing an AGI study from 1988, also says that 75% of women "say having a baby would interfere with work, school, or other responsibilities." Because these percentages include women each giving multiple reasons, it would be erroneous to select a single reason as decisive in these decisions to obtain abortions. In the 1988 AGI survey, only 21% cited economics as the most important reason, and only 13% cited issues with the mate as the most important reason. In data reported by the state of Minnesota, from 1998 to 2003 only 25.8% of women reported economic reasons and only 18.9% reported relationship issues or single status as a reason. In data reported by the state of South Dakota, from 1999 to 2003 only 28% of women cited economic reasons for their abortions.

3. Connection between unemployment and abortion rates: In most of the articles reviewed here, Stassen claims a correlation between unemployment and abortion rates. Most of his cited data are contradicted by AGI and BLS figures:
   • Women's unemployment increased from 6% in 1973 to 7.6% in 1980, decreased to 5.5%, then "partway up briefly" to 7% in 1992, then dropped "nicely" to 4.5% under Clinton, then rose "above 6%" under Bush: This misrepresents the data. The graph below shows abortion rates per thousand women from the AGI and female unemployment from the BLS. It also shows Stassen's summary of the behavior of unemployment rates from 1973 to the present. Notably, Stassen misrepresents unemployment rate behavior from 1973 to 1985, ignoring two sharp peaks in female unemployment that are poorly correlated with abortion rates.
     

     In fact, female unemployment rose from 6.0% in 1973 to 9.3% in 1975, fell to 6.8% in 1979, then rose again to 7.4% in 1980. It then rose again to 9.4% in 1982, fell to 5.4% by 1989, then rose to 7.0% by 1992. While it did drop to 4.1% during the Clinton administration, it did not reach 6% under Bush (at least in regard to annual averages): female unemployment was 4.7% in 2001, 5.6% in 2002, 5.7% in 2003, and 5.4% in 2004.

   • Abortion rates have moved "in tandem with" women's unemployment for 30 years: The correct unemployment figures, it turns out, are poorly correlated with abortion rates. In the late 1970s abortion rates were rising while female unemployment was falling; the end of the rise in abortion rates corresponds to rising unemployment. The post-1980 decline in abortion rates stalled from 1988 to 1990, at the same time as an 18-year low in female unemployment. Finally, the current peak in unemployment does not correspond to an increase in abortion rates.

4. Connection between marriage and abortion rates: Stassen claims more abortions result from lack of a "reliable husband or mate", which in turn he claims results from high unemployment. He specifically cites reduced marriages in his 16 chosen states. However, the decline in marriages and marriage rates has been mostly ongoing since the early 1980s. Marriage indicators do not correlate with economic indicators such as unemployment. Abortions fell during the largest rise in unemployment since Roe v. Wade--which occurred in the early 1980s. There are other social factors at work here, which Stassen ignores.

   The graph below shows abortion rates and marriage rates:

   

   Marriages have been declining consistently over time; during most (but not all) of this period, abortions have declined as well. Unemployment, in contrast, has risen and fallen repeatedly in this period. In addition, state-to-state data show no correlation between marriage rates and abortion rates.

5. Connection between income and abortion rates: In S1, Stassen notes that the federal minimum wage has been constant for seven years and implies a connection between real minimum wage/real income and abortion rates. The graph below shows minimum wage in 2005 dollars (i.e. minimum wage adjusted for inflation, using the BLS's CPI index) and abortion rates. The most significant decline in the real minimum wage was from 1978 to 1989; for 9 years, from 1980 to 1989, the minimum wage remained at $3.35. But this period corresponds to a decrease in abortion rates.
   

6. Connection between health insurance coverage and abortion: In S1, Stassen says "Since 5.2 million more people have no health insurance now than before this presidency--with women of childbearing age overrepresented in those 5.2 million--abortion increases." U.S. Census Bureau data indicate that there were 3.8 million more uninsured people in 2002 than in 2000. However, this increase did not particularly include women of childbearing age. Of the 3.8 million, 2.5 million were men and 1.3 million were women. The percentage of people without health insurance was 13.9% in 1990, 15.4% in 1995, 14.2% in 2000, 14.6% in 2001, and 15.2% in 2002. Women are underrepresented (13.9% without insurance in 2002 vs. 16.7% for men), as are children (11% without insurance for those 11 years old and younger). The percentage and number of people without health insurance has risen since the late 1990s, but this rise began before the current presidential administration and has yet to reach the percentage uninsured in 1995 (according to data available now, which only covers through 2002).

7. Connection between infant mortality and government-supported health care: In S5, Stassen states that infant mortality increased in 2002 for the first time "ever since 1940". He further attributes this to "decreased support for mothers and babies and decreased healthcare."

   According to the CDC, infant mortality did indeed rise from 6.8 infant deaths per 1000 live births in 2001 to 7.0 per 1000 in 2002. However, the downward trend since 1940 is not unbroken: in 1958, infant mortality briefly rose. The CDC report indicates that analysis of factors for this increase is pending, but does indicate that the rise is "concentrated in the neonatal period (less than 28 days), particularly the first week of life." (p. 12). Percentage of preterm births and low birthweight births also rose. They report no change in late fetal death rates, no change in the sum of late fetal death rates and early neonatal deaths, and no change in postneonatal death rates.

8. Discussion of economic policies/conditions by presidential adminstration: Stassen makes several claims which might have come from the Democratic National Committee rather than the U.S. Dept. of Labor. These are numerous and have been addressed by others in various forums. A few examples follow.

   He states in S1 and SK2 that unemployment rates "increased half again" in the last three years. In reality the unemployment rate rose from 4.1% in 2000 to 5.7% in 2003 (a 39% increase), then fell to 5.4% in 2004. Data thus far for 2005 indicate a continuing decline.

   Another statement made in S1 and SK2 is "Not since Hoover had there been a net loss of jobs during a presidency until the current administration." This was a specific claim by the Kerry campaign, but is at best misleading. The number of employed people rose in President G. W. Bush's first term, and had shown a net increase at the time of Stassen's claim.

9. Connection between abortion rates and social/economic factors at the state level: In several cases Stassen claims a correlation between state-by-state variations in abortion rates and variation in other measures. He gives state-level data on marriages for states in his sample of 16 states, and in S3 he describes Minnesota as a "state [that] invests in education, and therefore the unemployment rate is among the nation's lowest, and the abortion rate is the very lowest."

   A simple statistical test for correlation between two variables is the correlation coefficient. If one variable is proportional to the other (and likewise increase), the CC value is 1. If the relation is linear but one decreases as the other increases, the CC value is -1 (this is anti-correlation). CC values near 0 indicate no correlation.

   This test can be applied to state-level data to test for correlation between abortion rates and the assorted variables Stassen claims are correlated to abortion rates. These results are given below:

   CC between abortion rate and:	AGI, 2000, occurrence data	CDC, 2001, occurrence data	CDC, 2001, residence data	remarks
   murder rate (per 1000 pop.)	0.64	0.44	0.3	some correlation
   unemployment rate (civilian, %)	0.23	0.25	0.3	slight correlation
   male unemployment rate (civilian, %)	0.19	0.24	0.32	slight correlation
   female unemployment rate (civilian, %)	0.23	0.22	0.24	slight correlation
   marriage rate (per 1000 pop.), excluding NV	-0.32	-0.14	-0.17	slight correlation
   fraction of pop. in poverty (%)	0.04	-0.01	-0.06	no correlation
   fraction of pop. without health insurance (%)	0.07	-0.04	-0.02	no correlation
   marriage rate (per 1000 pop.)	0.1	0.14	0.18	slight correlation with more marriages
   public education spending ($ per student)	0.59	0.54	0.59	some correlation with high spending
   disposable per capita income ($)	0.65	0.58	0.6	some correlation with high income

   Stassen claims a correlation between homicide rates and abortion rates; there is some correlation, less for data by state of residence than by state of occurrence. There is slight correlation between unemployment rates and abortion rates. For the other variables examined, the relations Stassen claims either do not exist at all or are the opposite of what he claims. There is no correlation of abortion rates to fractions of the population either in poverty or lacking health insurance. Higher abortion rates are somewhat correlated with higher education spending or higher disposable per capita income--the opposite of what Stassen claims. This data was available to Stassen.

   Regarding Stassen's claim of correlation between state education spending and unemployment, the CC value is 0.03, indicating no correlation. For his claim of correlation between male unemployment and marriages, the CC value is 0.08, indicating no correlation (the slight positive value indicates rising marriages with rising unemployment).

10. Comparison to abortion rates and government policies in Belgium and the Netherlands: Stassen errs in citing abortion rates and social policies in other countries in support of his thesis. He cites Belgium and the Netherlands, claiming their abortion ratios are lower because of "strong economic support as well as health insurance" for mothers. However, many other countries have similar social policies and nonetheless have abortion ratios comparable to the U.S. Belgium and the Netherlands are distinguished by having a mandatory waiting period for abortions. In other words, legislative restrictions on abortion rather than socialist economic practices are best correlated with lower abortion rates.

Stassen's statements regarding his pro-life position

In S1 and S2, Stassen describes himself as "consistently pro-life." O'Bannon and Hussey, in the first NRLC article, pointed out that Stassen signed "A Call to Concern" in 1977, a document they said "expressed support for the Roe v. Wade decision and affirmed that 'abortion in some instances may be the most loving act possible.'" Stassen responded strongly in R3: "I did not sign a statement in 1977 supporting Roe V Wade; along with very large numbers of Christian ethicists, I signed a statement supporting academic freedom for Christian ethicists and moral theologians who take varieties of positions on these issues, and who were under pressure in some schools. I do not appreciate the personal attack."

The full text of "A Call to Concern" was posted by J. Taylor at Between Two Worlds. There is no reference to academic freedom in the document (although the final statement is "In the long run, the true test of ecumenical authenticity is the ability to sustain dialogue and friendship in spit of very sharp disagreements on matters of substance."). It does state "We support the Supreme Court decisions of 1973 which had the effect of removing abortion from the criminal law codes." While it does state that abortion "cannot be blandly legitimized," the focus of the document, including its five points, is to criticize the pro-life position. The statement, signed by Stassen, quite clearly supports the Roe v. Wade decision.

In several articles (S1, SK2, S3, S5, and S6), Stassen gives a personal family story. As relayed in S1, "For my family, 'pro-life' is personal. My wife caught rubella in the eighth week of her pregnancy. We decided not to terminate, to love and raise our baby. David is legally blind and severely handicapped; he also is a blessing to us and to the world." This is commendable and is a testimony to the value of human life.

It may be noted that the context of this decision was quite different than that today. In S6 he indicates that his son was born in 1967. In 1967, abortion laws had been liberalized in only three states: Colorado, North Carolina, and California. Callahan reports that a 1967 survey found the fraction of hospitals permitting abortion in cases of maternal rubella were 79% among non-Catholic private hospitals and 66% among public hospitals. Clearly, there was not the prevalence of support for abortion in such as case then as today. This is not to say that the Stassens would have made any other decision; their decision was from personal conviction (S3). This is only to note that the context of this decision is different than what a reader of S1 or SK2 might conclude.

Conclusions

Stassen's quantitative analysis gave an increase, rather than the decrease found by the AGI and CDC in far more complete and robust analyses. That his approach was unreliable results from incomplete data collection, errors in data entry, use of a sample that is both small and biased, and incorrect statistical treatment. Many of these issues were pointed out immediately, but Stassen has responded by affirming his approach and minimizing the significance of the AGI and CDC findings.

His broader thesis, however, is even more flawed. The minimal data that Stassen offers to support his linkage of government socio-economic policies and abortion rates are often misrepresented. The data in general refute his claimed correlations. More importantly, correlation does not prove causality: it is insufficient scientifically to merely demonstrate a correlation. Prove of a cause and effect relation requires demonstrating the mechanism. An important step here is use of controlled studies. A major flaw in Stassen's thesis is that he ignores the multitude of other factors that influence abortion rates. Scholarly studies would control for such factors by analyzing multiple subsets. Stassen has made no reference to controlled studies.

Stassen's claims not only lacked peer review, they were offered in what he describes as an opinion piece. For these claims to garner nationwide credibility among both the news media and politicians is quite telling. Importantly, no academics experienced with abortion statistics rose up to offer scientific analyses in his support. When such analyses did arrive, from the CDC and AGI, they weighed against Stassen's conclusions.

How Stassen could be so off the mark is suggested by a statement in S3: "The data I analyzed are simply a supplement to what anyone who assists pregnant women should know intuitively." It appears, particularly in light of Stassen's reaction to his critics, that his conclusions were predetermined. Some of his statements suggest a greater concern with a particular outcome rather than with a factual outcome. In S8 he attacks his critics for disputing his results: "Had my estimate... turned out to be right, it would have put significant pressure on the Bush administration... You would expect consistently pro-life advocates like me to hope I was right, and the increase the incentive and the pressure to reduce abortions... It raises the question whether their real loyalty is pro-life, or whether their priority is partisan politics." Observers, both expert and non-expert, do and will continue to disagree on the appropriate policies to reduce abortions. Such disagreements should not cloud the analysis of relevant data.

Sources--background