One of the jobs of nonpartisan analysts is to keep the parties honest. Partisans have a tendency to talk themselves into certain opinions, and there are enough data out there to make any case they wish. Nowhere is this more obvious than in the expectations game.
Over the past few months, both Democratic and Republican officeholders have noted that midterm elections usually result in losses often sizable losses for the presidents party. Democrats make the assertion so that theyll be able to claim victory if the party suffers a small net loss next November, while Republicans cite the trend as a recruiting tool and to boost party morale.
Just recently, a Republican Congressional candidate told me that he expected his party to make big gains next year because the average midterm loss for a presidents party is 30 seats. That number has also been cited by Democratic Congressional Campaign Committee Chairman Chris Van Hollen (Md.) in his effort to set a bar that he certainly can beat.
In fact, averaging midterm results over the past 100 or 150 years, or even the past 50 years, to set expectations for 2010 makes no sense. Averages are meaningless when you have a wide range of outcomes, as there have been in midterm elections since 1960, a reasonable place to start examining recent midterm trends.
Over the past five decades, there have been 12 midterm elections. Five of them resulted in large losses of at least 20 seats by the presidents party: 2006 (30 seats), 1994 (52 seats), 1982 (26 seats), 1974 (49 seats) and 1966 (47 seats).
Another five elections resulted either in small losses or gains by the presidents party 2002 (+8 seats), 1998 (+5 seats), 1990 (-8 seats), 1986 (-5 seats) and 1962 (-4 seats). The remaining two outcomes were between the two extremes, with the presidents party losing 15 seats in 1978 and 12 seats in 1970. (Midterm numbers come from Vital Statistics on Congress: 2001-2002 by Norman J. Ornstein, Thomas E. Mann and Michael J. Malbin.)
Clearly then, midterm results over the past 50 years have generally fallen into one of two ranges either small, single-digit changes (even sometimes in the favor of the presidents party) or large swings of at least two dozen and sometimes close to four dozen seats against the presidents party.
Averaging the outcomes of these dozen elections exaggerates the importance of the years with the biggest swings (1994, 1974 and 1966), ignores the bimodal distribution of election outcomes and produces a number that doesnt come close to reflecting what has actually happened in the elections.
To be sure, the general trend that the presidents party loses seats in midterms certainly is worth noting, as is the fact that in only two of the past six midterms has the presidents party suffered anything more than
single-digit losses. Indeed, in two of the past three midterms, the presidents party has gained seats. So we should expect Democrats to gain a handful of seats, shouldnt we? Obviously not.