Are the off-year gubernatorial elections in two East Coast states really the weather vanes that tell us which way the political winds are blowing, also known as the harbingers, bellwethers, omens, tea leaves and augurs of the political future? (Gotta get a new thesaurus.)

The politically-obsessed among us are already aware that in Tuesday's primary election in Virginia, state Sen. R. Creigh Deeds upset bigger-name Dems (especially Clinton pal Terry McAuliffe) to win the nomination for governor. Mr. Deeds (that's a Frank Capra reference) will face off in the fall against the Repub. nominee, Attorney General Robert F. McDonnell, who narrowly beat Deeds for that office in 2005.

Stories about it inevitably invoke the notion that this is one of two 2009 races -- the other is the New Jersey guv race -- that pundits use to measure the direction of the political winds.

At least among the cliche prone political analysts and writers, this is a common belief. You'll find the actual word "harbinger" (and let's face it, this word doesn't come up randomly all that often) in this Politico piece and this one from National Journal/Hotline. The Wash Post simply headlined its update on the two races: "Close Race in Virginia, New Jersey May be Indicators for 2010."

Vin Weber said, when I covered him at the Humphrey recently, that those who worry about the depth of the Republican hole should look past the attention that Dems are gleefully spending on Rush Limbaugh and Dick Cheney and follow those two guv races as evidence of the GOP's "political viability."

This happens every four years with the same two races. By two quirks of the political calender, in the year before a midterm congressional election, Virginia and New Jersey are the only states that hold gubernatorial races. I guess if any states could really be bellwethers for the nation, those two are fairly well-qualified, since both are swing states -- at least in gubernatorial politics. In the last 10 guv races, always in those pre-midterm years, both states produced five Dem winners and five Repubs.

But I've always been skeptical of this harbinger stuff, based on the assumptions that guv races are quirky, dependent on local conditions and the political strength of individual candidates, that no state or two can really reflect the nation's politics, that everything can change in the year after the bellwether has rung, and that political pundits are so obsessed with fortune telling that they might uncritically accept every possible opportunity to read any ol' tea leaves that come along.

And how silly must it be, methought, to look at two guv races for evidence of the national political wind, when the the two parties may easily split the two races, which would seem to augur nothing (or augur whatever the tea-leaf reader chooses to see)?

So this year, as the harbinger references start to pile up I decided to see how well those two odd-year guv races have foreshadowed the midterms of the following year. My conclusion is that the bellwetherness of the two guv races is greatly overrated.

First of all, the two states have cooperated over the past five cycles. One party or the other has won both guv races in VA and NJ. In three of those five cases, the same party has won -- in other words, has picked up seats in both houses of Congress -- the following year.

Outside of the five-day weather forecast, getting something right three out of five times would hardly be a strong case for reliable prediction. But it gets worse. The party that controls the White House almost always loses seats in the congressional midterms. The exceptions to that have been few over many decades. So, to really be impressed with the augury power of the NJ/VA guv races, you'd be looking for it to predict either really big midterm wins, or surprising midterm wins by the party in the White House. On that score, the NJ/VA weather vane has pointed in the correct direction in only two of the past five cases, the only two really big midterm landslides in the last two decades, namely:

• 1994 when the Repubs picked up eight Senate seats and 54 House seats and gained control of the House for the first time in 40 years, and that one was preceded by a 1993 Republican sweep of the NJ and VA guv races; and
• 2006 when the Dems picked up six in the Senate, 31 in the House, won control of both houses and a majority of the nation's governorships and state legislatures, and that one was preceded by a 2005 Dem sweep of VA and NJ.

So far, so good for the augurability of the NJ/VA factor. But after that, the test turns south, and I don't mean toward Virginia.

• In 1990 (following a 1989 Dem sweep of NJ/VA), the Dems gained just one Senate seat and seven in the House. This was in the middle of George H.W. Bush's term. By the historical standards of midterms, where the president's party is expected to lose, this rates as a Repub success where the weathervane would have predicted a solid Dem pickup.
• 1997-98: The 1997 "harbinger" elections in VA and NJ both went to the Repubs. The midterms were in the middle of Bill Clinton's second term, after his 1996 landslide over Bob Dole. So both the historical pattern and the NJ/VA weathervane would have predicted a good Repub year. Instead, the Senate experienced no change, and the Dems managed to gain five seats in the U.S. House. Strike two against the weathervane.
• 2001-2002: The Dems won both '01 guv races. George W. Bush was in the middle of his first term. But it was the Repubs who won the midterm, with +8 in the House and +2 in the Senate. Waterloo for the weathervane.

If, for reasons that surpasseth understanding, you are desperately inclined to believe in the NJ/VA harbinger, you might say that five cases is not enough to judge. Next problem: In the four cases preceding the five I just cited, New Jersey and Virginia did not agree with each other on the future direction of the national political wind, with one of them electing a Dem and the other a Repub. Without bothering to go over who won those midterms (except to note that the real historical pattern -- the president's party losing congressional seats -- held true in every case), it's hard to see how a mixed result from the two weathervanes could win them any credit for predicting the midterms.

If you want to go back further than 1973, you'll have to wait for my forthcoming book on this fascinating topic. (Just kidding. I'm getting sick of this piece too.)

What I see shaping up heading into 2010 is the strength of the historical pattern of presidential parties losing ground, versus the actual races that are shaping up, where every neutral analyst sees plenty of potential Senate pickups for the Dems, and very few for the Repubs. That's the real tension over which 2010 will be decided and anyone who thinks they know in mid-2009 -- or even in November 2009 when New Jersey and Virginia will weigh in -- is kidding himself.