Adaptive Retrospective Citizens Partisan Cycles
Divergence, therefore, is an empirical matter. Empirically, partisan economic policy/outcome differentiation is obvious. Tufte (1978), for example, finds US party platforms contrast more on economic and labor issues than on most others. Democrat and Republican voters divide similarly, though less sharply. Inflation and unemployment concerns, particularly, are highly cyclical and common to all, but persistent partisan differences manifest, mirroring the socioeconomic characteristic of each party’s constituency. Hibbs establishes left/right priorities most thoroughly, stressing relative unemployment/inflation aversion. Hibbs (1987a) shows, exhaustively and indisputably, that lower ends of occupational, income, and societal hierarchies face greater, more cyclical unemployment risk, and that tax-and-transfer systems only partly mitigate this risk.
While unemployment’s aggregate costs are large and obvious, Hibbs (and most others) find no evidence that inflation, short of hyperinflation and distinct from relative prices and inflation variability, harms almost any real outcome, including average income tax rates; aggregate real revenues, growth, investment, or savings; or the non-residential/housing investment mix. The only appreciably deleterious inflation effects appear in profitability, capital, and stock returns. Therefore, objectively, to the extent that these effects of inflation are felt more by upper classes, and to the extent that upper classes face less unemployment risk, they will have relatively more dislike of inflation than unemployment than do lower classes. Hibbs’s partisan theory, in fact, requires only that this ratio of unemployment–inflation aversion among lower classes exceed that ratio among upper classes. Indeed only the ratio of perceived, and not necessarily objective, aversions matters, which Hibbs thoroughly demonstrates, estimating that Democratic voters penalize incumbents 1.1 times as much for 1 per cent unemployment as for 1 per cent inflation, substantively and statistically significantly greater than Republican voters’ penalty ratio of .65. Thus, different voter groups suffer disproportionately from unemployment or inflation; public perceptions reflect this objective difference; and incumbents’ electoral approval follows suit, yielding differing partisan incentives to combat unemployment or inflation. More generally, party platforms differ on a range of economic issues; voters recognize and act on these differences; and parties enact policies accordingly. Therefore, left parties seek higher growth/employment and will accept higher inflation if need be to get them; right parties behave oppositely.
(Left parties will also expend greater equalization efforts.) Hibbs assumes monetary, fiscal, and other policies have sizable short- to medium-term4 impact, so these policy differences should manifest in outcomes also. Supportive evidence is plentiful relative to that for electoral outcome cycles. Hibbs (1977, 1987a) estimates 1.5–2 per cent higher unemployment and 5.3–6.2 per cent lower real growth under US Republican administrations than under Democratic. Democrats also contributed 60± per cent of the 1948–78 reduction in 20/40-ratio income inequality. Beck (1982) finds these estimates inflated by about 1/3, but qualitatively concurs. Hibbs (1987b), Paldam (1989), and others find similarly in broader OECD samples. Alesina and his various colleagues (see n. 2 above) also find evidence of partisan cycles in the USA, as well as in the OECD countries.
Several of these studies find context-conditional partisan cycles. Considerable consensus exists, therefore, about the role of partisanship in cycles of worsening nominal outcomes (like inflation) and improving real (unemployment, growth, etc.) and distributional outcomes under left governments, in US and comparative data. (Clark and colleagues5 dissent, finding context-conditional cycles that favor electoral more than partisan models.) An enormous literature analyzes empirical evidence of cycles in partisan policy. Imbeau, Petny, and Lamari (2001), for instance, meta-analyze 37 out of 600 partisanpolicy studies that address economic policy, spanning welfare, education, health, social security, privatization, intervention, public employment, spending, revenue, debt/deficit, etc., yielding 545 coefficients of these, 72.5 per cent sign intuitively (e.g. left parties supporting policies favorable to their traditional constituencies), with 24.8 per cent significant at p ≤ .10; 26.6 per cent have wrong sign, 8.3 per cent significantly; 0.9 per cent report no relation. This is a fair record, particularly when the varying and sometimes simplistic structure of the component studies is considered. The strongest partisan effects emerge from more sophisticated analyses, post-1973 samples, and regard government size: revenue, spending, employment (especially), or social welfare (less so). These results suggest that partisan cycles, like electoral cycles, follow a Ramsey Rule: all/most policy tools are used to meet partisan ends, although certain tools are preferred, and the extent and the mix of usage of policy tools exhibit strong context dependence. Summarizing, empirical research demonstrates that nominal outcome effects exhibit partisan cycles most strongly, although partisan cycles in distributional outcomes are also evident and even real-outcome partisan cycles receive moderate support. Broad partisan policy cycles are also consistently and strongly evident, with strongest evidence for public employment but also some for spending and revenue. Finally, partisan cycles are found more consistently in social or welfare policies, tax structure, and monetary policy, than in fiscal policy.
Naive left-deficit, right-surplus arguments,6 for example, have least support. In sum, empirical evidence of partisan policy and outcome divergence, in the expected directions, recurs frequently. Estimates vary and standard errors are sometimes large (statistical significance low), however, which could suggest that partisan divergence is small as Clark and colleagues (see n. 5 above) contend. To us, the more likely explanation for this variation in estimated coefficients and these sometimes large standard errors in the empirical record are the insufficient sample sizes/variations, the inappropriate/inadequate controls, and the mis-specification of context-conditional relations as unconditional characteristic of most empirical studies. We emphasize especially the last of these because, in all cases, and perhaps especially in monetary and fiscal policy, the overall pattern of empirical results suggests highly context-conditional partisan cycles..
