The Party of the President Usually Loses Power in the States in Midterm Elections.—

The existing political science literature tends to emphasize the loss of seats in Congress in midterm elections, which it nicely documents. In a comment to appear soon in the Yale Law Journal (vol. 115, pp. 2611-2622), Steve Calabresi and I document the loss of state governorships in off-year and midterm elections. You can download a full copy from SSRN at the bottom of this linked page.

The backlash against the President's party in state races during a President's term is actually stronger overall than the coattail effect in the presidential election year. To be more specific, we find that four years after a party wins a presidential election, it holds on average three fewer statehouses than it had before it won the presidential election. Perversely, winning the presidency seems to lead very shortly to losing power in the states. Since 1932 there have been eight changes of party control of the White House (1933, 1953,1961, 1969, 1977, 1981, 1993, and 2001). In every instance but one, the party that seized the White House held more governorships in the year before it took office than in the subsequent year it lost the presidential election. The only exception is that in 1980, Republicans held four fewer governorships than they held in 1992, immediately before the Republicans were voted out of the White House. Similarly, of the eleven Presidents since 1933, every one except two, Kennedy and Reagan, left office with fewer governorships than his party had before he took office, and Kennedy served less than three years. Figure 1 shows this pattern.

Click to enlarge

The four-year pattern of a federal election cycle is shown in Figure 2.

Click to enlarge

If one looks at the pattern since 1960, in his first year in office, a President's party controls only one more governorship than the party had in the election year. Once he is in office, there is a backlash against the sitting President's party. On average, since 1960, by the third and fourth years of a four-year presidential administration, the President has lost four seats from his first year, thus losing not only that one "coattail effect" seat, but three more governorships as well. One sees a similar, but slightly stronger, pattern since 1936. If one looks at just two-term administrations since the 1950s, by the seventh year of the administration, the party winning the White House has nearly eight (7.6) fewer governorships on average than it had before it won the White House.

We attribute this effect to presidents being a lightning rod for everything that goes wrong, which tends to lead to lower presidential approval ratings:

Most Presidents leave office less popular than when they entered, with Ronald Reagan and Bill Clinton being the only exceptions since at least Dwight Eisenhower. Even the exceptions (Reagan and Clinton) suffered major congressional losses in their first midterm elections, at times when their job approval ratings were down substantially. Thus, the response of voters is to blame the President for whatever goes wrong and, probably as a result, to punish that President's party in midterm and off-year elections.

Further, the lower voter turnouts for off-year and midterm elections tend to give those who are motivated to vote a bigger influence, and dissatisfaction with the president in office is a potential motivation.

Most states—-and all of the big ones—-have moved to electing their governors in off-year and midterm elections, in part to decrease the influence of the presidential election cycle on the state elections. Our data suggest that this effort to avoid presidential influence has been unsuccessful.

Theoretically, our backlash/lightning rod thesis is in some respects consistent and in some respects at odds with the prevailing political science theories of midterm elections: (1) referendum theory and (2) surge and decline/coattail theory.

As has been common with my other collaborations with Steve Calabresi, my contribution to the Yale comment has been predominately empirical (and to a lesser extent, theoretical).

The 2002 midterm election was unusual in Congress because the party not holding the White House, the Democrats, lost some ground, though they gained three governorships. The perhaps analogous 1962 election (which took place only a week or so after the Cuban missile crisis) was also relatively kind to the party in the White House, but the Democrats in power did very poorly in the next midterm election, 1966. Usually, in a two-term presidency, either the first or the second midterm election (1938, 1950, 1954, 1958, 1966, 1974, 1982, 1994) is very damaging to the president's party in either Congress or the states or both.

Before commenting below, please at least skim the paper itself (it is only 12 pages long):

Steven G. Calabresi and James Lindgren, The President: Lightning Rod or King?, 115 Yale Law Journal 2611, 2611-2622 (2006) (which may be downloaded at the bottom of this linked page at SSRN).

James Lindgren (mail):

Powerblogs seems to be a bit balky today. Not only is the "hidden text" function not working for this post, but I thought I saw that a couple of readers had commented, but when I went to read them the comments had disappeared. I might have been mistaken about their having been comments, but if you did indeed post a quick comment, sorry for the glitch; please repost.
9.18.2006 2:29pm
spencere (mail):
Most Presidents leave office less popular than when they entered, with Ronald Reagan and Bill Clinton being the only exceptions since at least Dwight Eisenhower.

Wait a minute.

Bush I

That is three.


Still looks like three.

Looks like something pretty close to a 50-50 ratio.

I'm not sure how you can draw the conclusion you do.
9.18.2006 3:12pm
spencere (mail):
Sorry, forgot Carter.

But the point still stands. The sample looks too small to be significant.
9.18.2006 3:17pm
The President as a lightning rod is certainly supported by this.
9.18.2006 3:48pm
James Lindgren (mail):

I'd have to recheck, but I think we counted 7 to 2.

I believe that the following presidents left office less popular than when they entered:

Bush 1
Bush 2 (we explicitly treat Bush2 as if he left office in 2006, but he is very unlikely to recover his late Jan. 2001 ratings).

More popular:

Of course, this is not a sample; it is the universe that fits the criteria, so strictly speaking statistical significance is not involved. And, as we point out, both Reagan and Clinton lost a lot of support in their first midterm election when they were relatively unpopular.
9.18.2006 4:09pm
It would be interesting to see what happens in state legislatures during these same off-year elections. Governor's races can be the equivalent of Senate races, where personality can sometimes trump partisan trends. State legislative gains and losses would be a more pure indicator.
9.18.2006 5:01pm
Greedy Clerk (mail):
Although I don't think Lindgren and his co-author are being misleading, I think the use of Clinton and Reagan as the examples of Presidents having higher approval ratings at the end of office is easily explainable -- they are the only two Presidents on the list (except for Ike) who served a full two terms. The others were either not reelected (Ford, Carter, Bush), chose not to run for reelection (LBJ) or resigned from office in disgrace (Nixon). (JFK should really not be included as his Presidency ended very prematurely and in tragedy.) Thus, assuming W. Bush leaves office less popular than when he came in (which is still an assumption which might not hold), he would be the historical aberration, not Clinton and Reagan.

As to Ike, he is really not a President of the modern television era. Further, his popularity was so high on entering office that it would be hard to sustain that.

Of course, as I think the authors note, the "sample size" is necessarily not going to be statistically sound because there are only so many Presidents in the last 50 years, and each of them obviously served at different times and faced crises and events (many of which beyond their control) which influenced their popularity at the end of their service.
9.18.2006 5:12pm
Greedy Clerk (mail):
Also, Ike was President before the Voting Rights Act which dramatically changed the electorate in key portions of the country.
9.18.2006 5:14pm
steven lubet (mail):
"The 2002 midterm election was unusual in Congress because the party not holding the White House, the Democrats, lost some ground."

Perhaps this anomaly is due to the fact that the Republicans, though holding the White House, had not actually won the 2000 presidential election (even forgetting the florida fiasco, Gore beat Bush nationally by 500,000 votes). Hence, there weren't any vulnerable "coat tail freshmen" whose seats were at risk. The Democrats did pick up a few governorships in 2002, thus demonstrating the validity of the Lindgren/Calabresi thesis.
9.18.2006 5:48pm
John Herbison (mail):
With one exception, the greatest pickups in Congress by the party out of power have come six years after the presidency changes parties and the president's party has been elected to a second term. Republicans gained significantly in 1938, Democrats in 1958, Republicans in 1966, Democrats in 1974 and 1986. The exception is 1998, which was skewed by the Republicans' having made great gains during the mid-term election during President Clinton's first term.
9.18.2006 7:01pm
Steven Lubet:

2002 may have been an aberation, but to follow up that up with the 2004 elections where the Republicans picked 3 more seats in the House and 4 in the Senate argues against that interpretation.
9.18.2006 8:22pm

The others were either not reelected (Ford, Carter, Bush), chose not to run for reelection (LBJ) or resigned from office in disgrace (Nixon).

Why in the world would you choose to treat presidents who were not reelected as a aberration from "not as popular when they left office as when they entered?"

Isn't not getting reelected a pretty solid correlation to not being popular?

For that matter, since popularity is at least supposed to be based on one's performance in office, "leaving in disgrace" again counts as a pretty clear indication of lost popularity, doesn't it? Getting pitched out of office shouldn't count? Yikes.

And choosing not to run for reelection should hardly exempt one from being a data point. Do you really assume that LBJ would have chosen not to run after one term if he had high approval ratings and a successful Presidency?
9.19.2006 2:12pm
Rich Rostrom (mail):
There could be another factor. The party which wins the Presidency may be presumed to have a stronger candidate, the pick of its crop of new faces. This suggests that prior to the Presidential success, there was a surge in that party's success in state races, i.e. governorships. Thus by the next mid-term elections, the "party in power" would be exposed (holding more governorships) and probably facing the retirement of several of those governors who were "new faces" 4-6 years earlier.
9.20.2006 3:41am