¹ Jackman, Robert, ‘Political Institutions and Voter Turnout in the Industrial Democracies’, American Political Science Review, 81 (1987), 405–423CrossRefGoogle Scholar; Jackman, Robert and Miller, Ross, ‘Voter Turnout in the Industrial Democracies during the 1980s’, Comparative Political Studies, 27 (1995), 467–492CrossRefGoogle Scholar; Blais, André and Ken Carty, R., ‘Does Proportional Representation Foster Voter Turnout?’ European Journal of Political Research, 18 (1990), 167–182CrossRefGoogle Scholar.

² Franklin, Mark N., ‘Electoral Participation’, in L. LeDuc, R. Niemi and P. Norris, eds, Comparing Democracies: Elections and Voting in Global Perspective (Thousand Oaks, Calif.: Sage, 1996), pp. 216–235Google Scholar; Blais, André and Dobrzynska, A., ‘Turnout in Electoral Democracies’, European Journal of Political Research, 33 (1998), 239–261CrossRefGoogle Scholar; Blais, André, ‘What Affects Voter Turnout?’ Annual Review of Political Science, 9 (2006), 111–125CrossRefGoogle Scholar. Studies also differ on the base on which turnout is calculated, some taking enrolled or registered voters, others the age-eligible population, building in non-registration to the non-voting category. However, of the two most recent studies showing the largest effects, one used a registration base and the other age-eligibility. See Franklin, ‘Electoral Participation’; and Norris, Pippa, Democratic Phoenix (Cambridge: Cambridge University Press, 2002)CrossRefGoogle Scholar.

³ Indeed, it may be possible to make further inferences about the supplementary effects of retention of an SMP component in a mixed PR system after change from SMP. On the main point here, while some tend to lump mixed systems together, there is a clear and major difference between mixed-member majoritarian (MMM) and mixed-member proportional (MMP) systems recognized and discussed by Massicote, Louis and Blais, André, ‘Mixed Electoral Systems: A Conceptual and Empirical Survey’, Electoral Studies, 18 (1999), 341–366CrossRefGoogle Scholar; and Shugart, Matthew and Wattenberg, Martin, ‘Mixed-Member Electoral Systems: A Definition and Typology’, in Matthew Shugart and Martin Wattenberg, eds, Mixed-Member Electoral Systems: The Best of Both Worlds? (Oxford: Oxford University Press, 2000), pp. 9–24Google Scholar. When MMP systems deliver high levels of proportionality, they fall into two categories, being both proportional and mixed. ‘PR systems’ cut across list, ordinal and mixed systems as proportionality is usually driven much more by district magnitude than ballot structure.

⁴ See, as a recent example of this argument, Blais, André, To Vote or Not to Vote: The Merits and Limits of Rational Choice Theory (Pittsburg: University of Pittsburg Press, 2000)CrossRefGoogle Scholar.

⁵ Banducci, Susan A. and Karp, Jeffrey A., ‘The Impact of Proportional Representation on Turnout: Evidence from New Zealand’, Australian Journal of Political Science, 34 (1999), 363–377Google Scholar.

⁶ Jackman, ‘Political Institutions and Voter Turnout in the Industrial Democracies’. This effect has recently been confirmed in the sense of ‘broad-based coalition governments’; see Karp, Jeffrey A. and Banducci, Susan A., ‘Political Efficacy and Participation in Twenty-Seven Democracies: How Electoral Systems Shape Political Behaviour’, British Journal of Political Science, 38 (2008), 311–334CrossRefGoogle Scholar; Banducci, Susan A., Donovan, Todd and Karp, Jeffrey A., ‘Proportional Representation and Attitudes about Politics: Evidence From New Zealand’, Electoral Studies, 18 (1999), 533–555CrossRefGoogle Scholar.

⁷ Miller, Raymond, ‘Coalition Government: the People’s Choice’, in J. Vowles, P. Aimer, Susan Banducci and J. Karp, eds, Voters’ Victory? New Zealand’s First Election Under Proportional Representation (Auckland: Auckland University Press, 1998)Google Scholar; Karp, Jeffrey A. and Bowler, Shaun, ‘Coalition Politics and Satisfaction with Democracy: Explaining New Zealand’s Reaction to Proportional Representation’, European Journal of Political Research, 40 (2001), 57–79CrossRefGoogle Scholar.

⁸ Jeffrey A Karp, Susan A. Banducci and Shaun Bowler, ‘Getting Out the Vote: Party Mobilization in a Comparative Perspective’; Vowles, Jack, ‘New Zealand: Consolidation of Reform?’ in Michael Gallagher and Paul Mitchell, eds, The Politics of Electoral Systems (Oxford: Oxford University Press, 2005), 295–312CrossRefGoogle Scholar, at p. 306.

⁹ See Karp, and Banducci, , ‘Political Efficacy and Participation in Twenty-Seven Democracies’Google Scholar.

¹⁰ The extensive literature on these matters and the related debates back and forth between qualitative and quantitative researchers are perhaps best traversed in Brady, Henry E. and Collier, David, eds, Rethinking Social Inquiry: Diverse Tools, Shared Standards (Lanham, Md.: Rowman and Littlefield, 2004)Google Scholar.

¹¹ Franklin, Mark N., Voter Turnout and the Dynamics of Electoral Competition in Established Democracies Since 1945 (Cambridge: Cambridge University Press, 2004), p. 207CrossRefGoogle Scholar.

¹² This article is not intended to be a comprehensive test of all aspects of Franklin’s theory with New Zealand data. For something closer to this, see Vowles, Jack, ‘Voting’, in Raymond Miller, ed., New Zealand Government and Politics, 4th edn (South Melbourne: Oxford University Press, 2006), pp. 315–326Google Scholar. But this analysis did not include the 2005 election.

¹³ Nagel, Jack H., ‘Voter Turnout in New Zealand General Elections 1928–1988’, Political Science, 40 (1988), 16–38CrossRefGoogle Scholar; New Zealand Electoral Commission, The New Zealand Electoral Compendium, 3rd edn (Wellington: New Zealand Electoral Commission, 2003)Google Scholar; New Zealand Electoral Commission, The New Zealand Electoral Compendium, 4th edn (Wellington: New Zealand Electoral Commission, 2006), see www.elections.org.nz/electoral-compendium-2005.htmlGoogle Scholar.

¹⁴ Banducci, and Karp, , ‘The Impact of Proportional Representation on Turnout: Evidence from New Zealand’Google Scholar.

¹⁵ Vowles, Jack, ‘Offsetting the PR Effect? Party Mobilisation and Turnout Decline in New Zealand, 1996–99’, Party Politics, 8 (2002), 587–605CrossRefGoogle Scholar; Karp, Jeffrey A. and Banducci, Susan A., ‘Political Parties and Voter Mobilisation’, in Jack Vowles, Peter Aimer, Jeffrey A. Karp, Susan A. Banducci and Raymond Miller, eds, Voters’ Veto: The 2002 Election in New Zealand and the Consolidation of Minority Government (Auckland: Auckland University Press, 2004), pp. 104–116Google Scholar.

¹⁶ This raises pertinent questions about the value and appropriateness of defining national competition in this way in a multi-party system under PR. The main objection to a two-party distance estimate is the potentially greater salience of bloc distance under multi-party PR. Yet in the New Zealand context, this poses problems in establishing the identification of centre parties in the process, and how to deal with uncertainties where small bloc-identified parties may be of questionable viability in terms of the electoral threshold. Given the equally problematic character of an alternative approach, a two-party estimate under MMP in New Zealand is justified here on pragmatic grounds. First and foremost, an expectation appears to have been established early that the major party with plurality over the other has the greater claim to govern, almost regardless of the performance of its allies. This is reinforced by the pivotal role of the centre parties that have almost without exception provided various gradations of support for the plurality winner. Support for these claims may be found in various places in Miller, Raymond, ed., New Zealand Government and Politics (Melbourne: Oxford University Press, 2006)Google Scholar, and Miller, Raymond, Party Politics in New Zealand (Melbourne: Oxford University Press, 2005)Google Scholar. Of course, the ultimate test lies in the analysis below: if significant expected effects are found for national competition as defined here, then, at least provisionally, this means of estimation can be defended as, at least, the best approximation at hand. This definition also correlates with Franklin’s average constituency competition variable at r = 0.89.

¹⁷ Miller, Warren E. and Shanks, J.M., The New American Voter (Cambridge, Mass.: Harvard University Press, 1996)Google Scholar; Lyons, W. and Alexander, R., ‘A Tale of Two Electorates: Generational Replacement and the Decline of Voting in Presidential Elections’, Journal of Politics, 62 (2000), 1014–1034CrossRefGoogle Scholar; Putnam, Robert, Bowling Alone: The Collapse and Revival of American Community (New York: Simon and Schuster, 2000)CrossRefGoogle Scholar; Blais, André, Gidengil, Elizabeth, Nevitte, Neil and Nadeau, Richard, ‘Where Does Turnout Decline Come From?’ European Journal of Political Research, 43 (2004) 221–236CrossRefGoogle Scholar; Vowles, Jack, ‘Voting’, in Raymond Miller, ed., New Zealand Government and Politics, 4th edn (Melbourne: Oxford University Press, 2006), pp. 316–326Google Scholar.

¹⁸ Franklin, , Voter Turnout and the Dynamics of Electoral Competition, pp. 59–64Google Scholar; Plutzer, Eric, ‘Becoming a Habitual Voter: Inertia, Resources, and Growth in Young Adulthood’, American Political Science Review, 96 (2002), 41–56CrossRefGoogle Scholar.

¹⁹ Vowles, , ‘Voting’, p. 320Google Scholar.

²⁰ Turnout on a validated basis is, of course, preferable to self-reported turnout, which tends to be overestimated in most surveys: see Karp, Jeffrey A. and Brockington, David, ‘Social Desirability and Response Validity: A Comparative Analysis of Over-Reporting Voter Turnout in Five Countries’, Journal of Politics, 67 (2005), 825–840CrossRefGoogle Scholar. While those misreporting tend to have attitudinal characteristics midway between voters and non-voters, this is still a source of bias. Unfortunately, validated turnout data in New Zealand has developed some flaws since 2002. See the Appendix for further discussion of this issue, and how its effects have been minimized in the analysis that follows.

²¹ Rosenston, Steven J. and Mark Hansen, John, Mobilization, Participation, and Democracy in America (New York, MacMillan, 1993)Google Scholar; Vowles, Jack, ‘Dealignment and Demobilisation? Nonvoting in New Zealand 1938–1990’, Australian Journal of Political Science, 29 (1994), 96–114Google Scholar; Vowles, ‘Offsetting the PR Effect?’

²² Data for each year of age for each cohort could be shown, but high sampling error in small cells makes it hard to read, so this statistical smoothing is the best approach.

²³ Missing are 1966, 1969, 1972, 1978 and 1984.

²⁴ Technically, these are age-period-cohort (APC) models, raising issues of ‘identification’ because of their origins in a linear association, each of these effects arising from a combination of the other two: see Mason, William M. and Frieberg, Stephen E., eds, Cohort Analysis in Social Research: Beyond the Identification Problem. New York: Springer-Verlag, 1985CrossRefGoogle Scholar. The ‘fix’ for this problem is a theoretically appropriate dropping of one of the three variables, or otherwise a linear restriction on one of them. Here, the links are weakened: the identification of dummy variables for cohorts and ‘period’ that are not simply mechanically related to the structure of the dataset. Instead, they are derived from theory and the specific hypothesis addressed. In fact, as there is only one ‘period’ dummy, this alone should solve the problem. (While this might be taken to under-specify theoretically relevant period effects, the addition of competition variables serves as a proxy for election by election differences, another route around the identification problem.) See, as an example of previous research in political science on this basis, Johnston, Richard, ‘Political Generations and Electoral Change in Canada, British Journal of Political Science, 22 (1992), 93–115CrossRefGoogle Scholar. Perhaps the most devastating criticism of APC models does not apply here – that over a long time series it makes no logical sense to control for a period where many respondents could not have experienced it, having either not been born at the time or having died before. But as the period effect here is a proxy for electoral system change, its logic is a counterfactual for the alternative electoral system: what might have happened if earlier respondents had experienced the conditions of PR, and what the most recent respondents might have done had they experienced the SMP system.

²⁵ Most specifically, and as indicated above, as there is only one ‘period effect’ dummy, this constrains other potential election-specific effects to 0.

²⁶ See appendix for details of the two alternative ‘historical competitiveness’ variables used. One of Franklin’s key claims was also tested in alternative models: namely, that new voters experiencing their first three elections will be significantly more affected by competitiveness. Various simple models on competitiveness, new voters and an interaction between them found no significant effects for the interaction in the hypothesized direction, either at time t or t−1. Testing another Franklin competitiveness indicator, no effects were found for the difference between the percentage of the largest party vote and 50 per cent, either as a main effect or as an interaction with competitiveness.

²⁷ All the election studies contain respondent codes for electoral districts. These data reflect the level of local competition experienced by each respondent at the election for which they were sampled, and the one before.

²⁸ STATA drops the main effect for the SMP system on grounds of collinearity when the boundary change interaction is included in this model, so this cannot be tested against social structure controls. Note that the model also excludes all but one of the elections at which turnout data were not validated, to some extent addressing doubts about the effects of employing both validated and non-validated data. Note also that the country margin at time t under SMP appears to have a negative effect, which is counter-intuitive. This is likely to be the result of collinearity with the district margin (r = 0.78) in the most restricted sample.

²⁹ It may reasonably be asked why party identification is not included in any of the models, particularly as a similar study of turnout over time in Britain relies on partisanship for much of its explanation of turnout decline there: see Heath, Oliver, ‘Explaining Turnout Decline in Britain, 1964–2006: Party Identification and the Political Context, Political Behaviour, 29 (2007), 493–516CrossRefGoogle Scholar. Party identification was included in alternate versions of the models, and its effect was to significantly boost their predictive power but without having major influences on the signs or values of the other variables. Levels of party identification in the pre-MMP samples averaged 67 per cent, and only 56 per cent under MMP. This is not consistent with cross-country evidence indicating that party preferences tend to be stronger under PR cited above and, on the surface, this could appear to account for the failure for turnout to increase under MMP. However, the addition of party identification to the models makes no difference to the electoral system effects estimated. Its effects are instead to mute somewhat the effects of new voters and competitiveness. Logically, however, in terms of the theoretical approach adopted here, one assumes that party identifications are affected by these variables, new voters being less likely to have identifications, and identifications tending to be affected by competitiveness and associated mobilization. Party identification is therefore an intervening variable that, if included in the models, acts to mask the effects of the more theoretically significant variables causally prior to it and, therefore, for present purposes is an inappropriate control.

³⁰ As noted earlier, while such social structure controls tend to be most strongly emphasized by alternative ‘resource’ theories of political participation, they can also be theorized in rational choice terms by way of access to information, differences in political knowledge (C), and efficacy (the latter influencing P).

³¹ It is worth noting that the incentive under the MMP system is for parties to shift their mobilization efforts from marginal to safe seats, as in their areas of strength parties should find mobilization costs lower and more potential to grow their vote by converting non-voters into voters. For an early investigation of this, see Denemark, David, ‘Campaign Activities and Marginality: The Transition to MMP Campaigns’, in Jack Vowles et al., eds, Voters’ Victory, pp. 81–100Google Scholar.

³² See the Appendix for the results of regressions of constituency competitiveness against constituency turnout at time t, and some at t−1, for the elections analysed here. From these data, it appears that the extent to which time t district competitiveness drives turnout may vary according to another contextual variable, the election winner. Labour party victories tend to be associated with larger effects, and National victories with smaller ones. But post-MMP, there have been three Labour victories, all with much lower effects than with Labour victories under FPP. ‘Party winner’ and an interaction between it and district competitiveness were included in alternate models, and neither proved statistically significant. One surmises that, for reasons of social structure, higher turnout is easiest to achieve in National-held safe seats, particularly when the party is heading for victory. The controls for social structure in the final model, therefore, take much of this into account. Where boundary changes do not complicate the comparison, the constituency-level data also confirm that the effects of district competitiveness tend to be stronger when measured at time t−1.

³³ Possible interactions between boundary changes and the effects of lagged district competitiveness on not voting were tested in the split SMP and MMP versions of Model V. This was to test the hypothesis that the almost flat line under MMP could be an artefact of the high proportion of boundary changes under the system, calling Figure 4 findings into question. However, the steepness of the relationship was the same with or without boundary changes under both systems, with the effects of change under MMP simply being stepped up at higher probabilities consistent with Figure 4.

³⁴ Franklin, , Voter Turnout and the Dynamics of Electoral Competition, pp. 61–8Google Scholar.

³⁵ Vowles, , ‘Dealignment and Demobilisation?’Google Scholar

³⁶ There are NZES panel data over up to four waves over the 1990, 1993, 1996 and 1999 elections, and again between 1996, 1999, 2002 and 2005. However, small Ns and high attrition make their use problematic, although it would no doubt be useful if these data were to be explored in further research.

³⁷ For example, Dalton, Russell, Democratic Challenges, Democratic Choices: The Erosion of Political Support in Mass Democracies (Oxford: Oxford University Press, 2004)CrossRefGoogle Scholar.

³⁸ Franklin, , Voter Turnout and the Dynamics of Electoral Competition, p. 168Google Scholar.

³⁹ Bean, Clive, ‘An Inventory of New Zealand Voting Surveys 1949–1984’, Political Science, 38 (1986), 172–184CrossRefGoogle Scholar; Vowles, Jack, ‘The New Zealand Election Study’, Political Science, 52 (2000), 150–157CrossRefGoogle Scholar; Vowles, Jack, Aimer, Peter, Karp, Jeffrey A., Banducci, Susan A., Miller, Raymond and Sullivan, Ann, Proportional Representation on Trial: The 1999 Election in New Zealand and the Fate of MMP (Auckland: Auckland University Press, 2002)Google Scholar; Vowles et al., eds, Voters’ Veto, and at http://www.nzes.org.

⁴⁰ The electorates dropped, identified by the codes used in the datasets, can be found at http://www.jackvowlesdomain.co.uk/dropped_electorates.html.

⁴¹ Note that electorate vote, not party vote shares, was used under MMP.

⁴² See http://www.indiana.edu/~jslsoc/spost.htm; and Scott Long, J. and Freese, Jeremy, Regression Models for Categorical Outcomes Using Stata, 2nd edn (College Station, Tex.: Stata Press, 2005)Google Scholar.