I am grateful for the opportunity to respond to Abramowitz, Alexander, and Gunning's (2006b) rejoinder to my critique of their article appearing in this same issue entitled “Don't Blame Redistricting for Uncompetitive Elections” (2006a). We should have had a scholarly debate that informed the profession and raised interesting questions in one of the widest read journals. Instead, Abramowitz, Alexander, and Gunning mischaracterize my arguments as if this was a political talk show debate, and it is telling that they never directly quote me in their rejoinder. The few sources cited in support of their methodology, upon close inspection, provide no support to their measurement choice or findings. I hope that fair-minded readers will carefully evaluate all material, including cited sources, to reach their own judgment. I conclude this article with some thoughts that I hope shed light on the low levels of electoral competition that Abramowitz, Alexander, and Gunning and I agree characterize current congressional elections.

Abramowitz, Alexander, and Gunning and I continue to have significant disagreements over the effect of redistricting on the number of competitive congressional districts. I take a measured approach with regards to my findings by stating: “A more proper conclusion is that both redistricting and underlying changes in the geographic distribution of partisans are contributing to the decline in the number of competitive districts” (2006a, 92, original emphasis). Thus, I agree with Abramowitz, Alexander, and Gunning's characterization of my findings that, “According to McDonald's data, the number of competitive districts declined by almost as much during non-redistricting cycles as it did during redistricting cycles” (2006b, 96). At first blush, we might agree.

However, Abramowitz, Alexander, and Gunning contradict their restatement of my findings in their next sentence: “Regardless of whether one uses the 45–55% standard or the 48–52% standard for identifying competitive districts, the decline in competitive districts between 1990 and 2002 that can be attributed to redistricting was approximately five districts” (2006b, 96). The juxtaposition of these two contradictory sentences is confusing and the authors do not describe the methods used to come to this conclusion, which is a much lowered effect than what I find. The authors might have interpolated the downward trend in district competitiveness over a decade and subtracted this from the change observed between a pair of redistricting years. From this, they might claim a reduced effect of redistricting on district competitiveness during a redistricting cycle. I apologize if I mischaracterize their approach, but it is consistent with their sentence: “But this assumes that no decline would have occurred in the absence of redistricting, which is clearly unrealistic” (2006b, 96). This approach might have led the authors to conclude: “… we find no significant change in the numbers of safe or competitive districts between 2000 and 2002” (2006b, 95), which clearly contradicts their Figure 1 (Abramowitz, Alexander, and Gunning 2006a, 88).

This approach, if it is the one the authors' take, fundamentally misinterprets the methods involved in my analysis and used by numerous other political scientists in their scholarly work on redistricting (e.g., Cox and Katz 2002; Gelman and King 1994a; Gronke and Wilson 1999; Swain, Borrelli, and Reed 1998). To isolate the effect of redistricting on district partisanship (and by extension, district competitiveness), the underlying partisanship of districts is measured before and after redistricting using the same measure, for example, the 1988 presidential vote within the 1990 and 1992 districts. The only change to districts is the distribution of partisans within districts, and thus there cannot be a change in district partisanship associated with anything other than rearranging district boundaries. Indeed, a close reading of two articles cited by the authors in support of their findings (Gopoian and West 1984, 1084; Glazer, Grofman, and Robbins 1987, 682) use a similar measurement of district partisanship to the one I employ. Abramowitz, Alexander, and Gunning provide no scholarly support for their measurement choice central to our disagreement.

Consistency is a reason why aggregating the same presidential election results within districts before and after redistricting is a preferable form of measurement to comparing different elections across a redistricting. Even if the reanalysis of Alvarez and Nagler (1995) by Lacy and Burden (1999)—who employ a model allowing for voter abstention—is wrong that Perot's presence in the 1992 election did not significantly reduce the national margin of victory between Bill Clinton and George H. W. Bush, the authors do not consider how the district level distribution of Perot's vote might confound their measurement. If there were no confounding effects, then the measure of competitive districts would be largely indifferent to the choice of presidential election analyzed. However, Abramowitz, Alexander, and Gunning (2006b, 96) find “148 rather than 164” competitive districts in 1992 using the 1988 rather than the 1992 presidential election results. Sixteen fewer districts or 10% of the measure's value is hardly “slight” (2006b, 96). Despite their downplay of the substantive differences, using the prevailing method of measurement within the scholarly literature significantly alters Abramowitz, Alexander, and Gunning's findings, both with regards to the 1990–1992 change and to the mid-decade change from 1992–2000.

Two scholarly studies cited by the authors in support of their findings narrowly focus on shifting partisans among incumbents' districts and the effect on incumbents' electoral success (Gopoian and West 1984; Glazer, Grofman, and Robbins 1987), rather than overall changes to district competition. Interestingly, Glazer, Grofman, and Robbins (1987, 698) identify “28 competitive seats which shifted” out of their defined competitive electoral range, though a full analysis of open and incumbent held seats is necessary to establish the relationship relevant to our discussion. These two articles cited by the authors do not support any claim regarding the relationship between redistricting and the total number of competitive districts since the unit of analysis of these studies is a subset of congressional districts: those with an incumbent.

Abramowitz, Alexander, and Gunning (2006b, 96) dismiss my findings by arguing that “much of my evidence” is “anecdotal” based on my “personal experiences” as a “consultant to pro-reform groups” and that it is therefore difficult to evaluate my arguments. The attack on the integrity of my scholarship is most disappointing. I note (McDonald 2006, 91) that my data on congressional districts—the data of primary contention—are drawn from Gary Jacobson and the Almanac of American Politics (Jacobson and I independently hand entered the Almanac's data). The only data source for Abramowitz, Alexander, and Gunning's is noted in Figure 1 as “Data compiled by authors” (2006a, 88). In their rejoinder, the authors do not contest my characterization that Jacobson provided their data (McDonald 2006, 91), which contradicts their claimed data source.

In my original critique, I attempted to faithfully replicate the authors' analysis based on personal communication I had with Alan Abramowitz, but we continue to have unresolved discrepancies. I am disappointed that the authors did not attempt the same courtesy when crafting their rejoinder as we might have been able to resolve our measurement discrepancies. Most notably, I find a decrease in the number of competitive districts in the 45–55% range between 1990 and 1992 using the 1988 normalized presidential vote within districts, from 150 to 146 (McDonald 2006, Table 1), while Abramowitz, Alexander, and Gunning find an increase, from 146 to 148 (2006b, 96). In fact, the authors contradict themselves. In their rejoinder they find “148 rather than 164” (2006b, 96) competitive districts using the 1988 normalized presidential vote rather than the 1992 presidential vote. Figure 1 in their first article (2006a, 88), however, shows for 1992 a number of competitive districts based on the 1992 normalized presidential vote that to my eyes does not exceed 160.

Perhaps the “anecdotal evidence” the authors refer to is my work for the Arizona Independent Redistricting Commission, which is not a “pro-reform group” but a redistricting authority. My work for the commission is extensively documented in the court case, Minority Coalition for Fair Redistricting v. Arizona Independent Redistricting Commission, referenced in a footnote in my article (McDonald 2006, 94). While anyone may obtain court documents necessary to replicate my case study of Arizona, all reports and transcripts of presentations I made to the commission are publicly available on the Arizona Independent Redistricting Commission's web site: www.azredistricting.org. The reports document my methodology and contain an analysis of district competitiveness for every map I examined for the commission.

Abramowitz, Alexander, and Gunning state, “[I] claim that Arizona provides a model of what can be accomplished by a truly nonpartisan redistricting commission [that] is not supported by the results of the 2001–2002 round of redistricting or by recent elections in that state” (2006b, 95). I never make this claim. My case study of Arizona illustrates the “rules that govern a commission constrain the number of competitive districts it can draw” (McDonald 2006, 93). Despite the lack of electoral competition, which I agree was experienced under the commission's maps, under court order I drew a state legislative map that “demonstrate(s) that it is possible to draw a large proportion of competitive districts, even in an unbalanced partisan state, if there is a political imperative to do so” (McDonald 2006, 93). A different state or a different set of rules would likely lead to a much different set of electoral outcomes. Indeed, reformers have tweaked the Arizona model. For example, Ohio's 2005 redistricting reform initiative moves competition higher on the rank ordering of criteria the commission would have to satisfy.

By mischaracterizing my case study of Arizona, the authors deflect my criticism that they classify all commissions as nonpartisan when analysis of their membership and rules reveals most tend to produce partisan or incumbent protection gerrymanders (McDonald 2006, 93). In their reply, the authors highlight their “finding that nonpartisan redistricting commissions have failed to produce increased competition” (2006b, 95). Yet, the authors still have not revealed which commissions they regard as nonpartisan, so the finding remains suspect.

My case study of Arizona demonstrates that nonpartisan commissions are not sufficient to produce competitive districts, as “nonpartisan” commissioners are constrained by context and institutional rules. Which institutional design best renders a preferred outcome is a difficult hypothetical question to answer, particularly when there are few or no institutions to generalize from. The method I use in my Arizona case study—essentially map simulations under hypothetical rules—is a methodology suited to address this question and has been used in a similar analysis of California's 2005 redistricting reform initiative, which also highlights a finding of increased district competition under an alternative redistricting institution (Johnson, Lampe, Levitt, and Lee 2005).