"Macro" Factors and Election Forecasts

by Chris Bowers, Thu Sep 21, 2006 at 04:04:30 PM EDT

Two weeks ago, Charlie Cook wrote about a generational gap among election forecasters. Basically, he wrote that older forecasters were more inclined to look at macro factors in House elections than were younger forecasters, who focused more on district-by-district analysis. I am not saying this to be a whippersnapper or anything, but when it comes to at least one macro factor, Presidential approval at midterms, I do have to seriously doubt it has a huge impact. Look at this:

2004. Seats won by Reps: 232. Bush approval index just before the election, according to Pollkatz: 47.4.
2002: Seats won by Republicans: 230. Bush approval index just before the election, according to Professor Pollkatz: 61.7

Pollkatz recorded a 14.3% difference in Bush's job approval from just before the 2002 and 2004 elections, yet in the House the results were nearly identical. Had it not been for Texas redistricting between the 2002 and 204 elections, Republicans probably would have only won 228 seats in 2004. As frustrating as it may be for our side this year, this isn't exactly the sort of evidence that makes one think macro factors play a large role in congressional elections. Or, at least, it doesn't lead to believe that Presidential approval plays a huge role in determining the makeup of the House. In fact, if you click through to look at this chart, there barely seems to be any correlation at all between Presidential job approval and the proportion of seats for Democrats in the House. For example:

In 1994, when Clinton's pre-election approval was just below 50%, Democrats won 204 seats in the midterm elections. However, in 1998, when Clinton's job approval was over 60% just before the midterms, Democrats only won 211 seats. Despite a swing of around 15 points in Presidential job approval, Democrats saw an improvement of only 7 seats in the House.
In 1982, when Reagan's job approval just before the midterms was just over 40%, Democrats won 269 seats in the House. In 1986, when Reagan's job approval was over 60% just before the midterms, Democrats won 258 seats. Despite a difference in job approval of roughly 20 points, Democrats won only 11 more seats.

This is not to say that the party of an unpopular, sitting President shouldn't worry about the impact of Presidential job approval on midterm elections. Certainly, every sub-50 score on that chart led to big losses for the party of the sitting President. In 1950, Democrats lost 29 seats when Truman's approval was low. In 1966, Democrats lost 47 seats when Johnson's approval was low. In 1974, Republicans lost 49 seats after Nixon resigned. In 1978, Democrats lost 15 seats when Carter's approval was low. In 1982, Republicans lost 27 seats when Reagan's approval was low. In 1994, Democrats lost 54 seats when Clinton's approval was under 50%. In all six cases, the party of the sitting President with a sub-50 approval rating (or the party of a President who had just resigned) lost seats in a midterm election. There just isn't any direct correlation between how many seats were lost and Presidential approval rating.

Charles Franklin currently estimates Bush job approval at 41.2%. This puts Bush as low as any other President heading into a midterm election from 1950 forward. This does indicate that Republicans will lose seats, but I think about 99% of the political community is expecting that to happen. What it does not tell us it that a landslide election is about to occur. Bush's job approval may be low, and Democrats may be poised to win seats, but I don't think that the macro factor of Presidential approval can be counted upon to forecast a wave election.

Last month, I sent up warnings about how the generic ballot is not necessarily a good macro factor to look at when determining how many seats Democrats will win in the House. Now, I am giving warnings about Presidential approval. The point I want to emphasize here is that while macro factors can give you a very generalized indication of how many seats Democrats will win in the house--namely, that they should gain seats--they are blunt instruments that fail when it comes to making precise projections. For that, I still believe the best solution is to look at district-by-district data. There may be other macro factors that are more important, such as the ones measured by Pew in their latest survey, but the information on those factors only goes back a few midterm elections. Maybe it is because I am young, but I will generally stick to a district-by-district analysis in my House forecasts, and only take the macro factors as a vague set of guidelines. I want to believe the macro factors that still point to a huge Democratic win, but I don't see enough evidence to be convinced of them--not by a longshot.

I will have an updated House forecast tomorrow.

Tags: election forecasts, House 2006, polls (all tags)

Comments

26 Comments

Sort by

permalink

Re: "Macro" Factors

This isn't about the House, but Kweisi Mfume's son appeared at a Steele campaign event.

Mfume is going to throw Cardin under the bus.

by Hesiod Theogeny 2006-09-21 04:16PM | 0 recs

permalink

Re: "Macro" Factors

Why do I get the same sinking feeling as the Townsend-Kennedy race in '02.

by adamterando 2006-09-21 04:26PM | 0 recs

permalink

The problem is the South

The South has basically been a monolithic block since 1878. First for Democrats, now for republicans. The amount of swing in other parts of the country varied over time, the northeast was solidly republican until roughly the new deal for instance. But the South has always been a one party region.

So when the south had a low population, from 1878 to roughly 1980, it was much easier to get big wave elections because if other regions were amenable to swinging between the parties, it would result in a large number of seats changing because the rest of the country made up a much larger share of the population.

Now though, the South has a very large share of the population. So the monolithic block plays a much larger role in mitigating the magnitude of a wave election.

This obviously isn't the only reason, gerrymandering is getting worse and worse, but I believe it's a major factor as to why wave elections will get smaller and smaller in the future as the south's population gets larger and larger. At least, that is, as long as conservatives control the levers of power.

That's why the 50-state strategy is so important.

by adamterando 2006-09-21 04:26PM | 0 recs

permalink

Only a small part of the South

I do think that while there's not going to be a huge wave from the South, there's a big difference between the Upper South (Tennesee, Kentucky,West Virginia) and the Deep South (Georgia, Missippi, Alabama). The Deep South is lost, the Upper South is not. We need to build the party everywhere, but I think that with the right investments we could have a real impact in the Upper South.

by ManfromMiddletown 2006-09-21 04:48PM | 0 recs

permalink

Re: Only a small part of the South

population wise I was focusing more on Georgia, Texas, and Florida.

by adamterando 2006-09-22 05:00AM | 0 recs

permalink

Texas

I think that Texas will trend Democratic over the next decade. It's two things, Texas is now majority minority, and that's only going to grow. Also there's internal immigration from California. I think that in the next 5-10 years we are going to see the transitions that shook California in the 1980's, and by 2020, Texas might be a Democratic state.

by ManfromMiddletown 2006-09-22 09:57AM | 0 recs

permalink

Re: Texas

I hope you're right.

by adamterando 2006-09-22 10:49AM | 0 recs

permalink

Re: "Macro" Factors

I don't think Presidential approval ratings are all that big of a factor in Congressional elections. I think it's about how people feel about Congress itself. Which of course is terrible - even worse than Bush's numbers.

I don't like these mindless statistics. They don't get us anywhere.

P.S. I think Mfume's son is an asshole. I hate it when Democrats like Mfume, Lieberman and Westly (in CA) become big babies when they lose primaries. I'm sure had Mfume pulled off a victory he would have expected the support of Cardin, but he's too much of a prick to reciprocate in kind.

by JackBourassa 2006-09-21 04:49PM | 0 recs

permalink

Re: "Macro" Factors

"I don't like these mindless statistics. They don't get us anywhere."

Thank you for that productive comment.

by Chris Bowers 2006-09-21 04:53PM | 0 recs

permalink

Surge years

Have you seen this NCSL article that takes on the issue of surge years in state legislatures, and basically agress with you.

The president of the NCSL (National Conference of State Legislatures) noted that some analysts point to the watershed years of 1994 and 1974 as evidence that there are these "macro" factors.

Today, state legislatures are almost perfectly divided between red and blue. Republicans have a majority in both chambers of the legislature in 20 states, and Democrats in 19. In 10 states, control is divided. Democrats have a symbolic advantage of holding a few more legislative seats nationwide than the Republicans, but that margin is less than 1 percent of America's 7,382 state legislative seats.

In 17 state senates, a shift of only three seats would put a new party in power. In 12 state houses, a shift of five or fewer seats would change control. In Colorado, Delaware, Iowa, Maine, Michigan, Montana, Nevada, Oregon and Tennessee, both chambers are vulnerable to change based on the numbers.....

The current state of national affairs has some things in common with other years of upheaval. The Jack Abramoff scandal certainly doesn't have the reach of Watergate, but it's an ethics scandal, and polls have shown that it has soured the public on elected officials. In the 1974 election after Watergate, Democrats won 628 seats, giving them control of a full 37 state legislatures. Eight were divided and the GOP had only four.

And war, too, has historically had an effect on elections. The Democrats gained 528 seats in the 1964 Johnson landslide election, only to lose them -- and then some -- two years later as Vietnam escalated. In 1966, Republicans seized back 762 state legislative seats.

I do suspect that there's going to be a tremendous impact at the state level. I think that in nationwide it's not unrealistic that disgust for Republicans as a seperate and distinct driver of votes will lead to a 5% shift from R to D, which means that any race the R's won by a margin of less than 10% is at risk.

Of course, there will be other factors at work, so that in places like the Deep South this won't matter as much. In places like Maine I think that Republicans in the state legislature are going to become exinct. I diaried the statistics on this on Kos.

In the 2004 election, 53 of Maine's 151 house seats were won by a margin of less than 10%, and 94 seats were decided by a margin of less than 20%. Maine is extremely competitive at the district level, having few of the landslide districts found in other states.

You have different forces macro forces at work here, but in states where you have multiple macro factors pointing towards D, that could lead to a huge swing. Like in Maine, where the long march to the left is being completed amidst a 67% disappoval rating for the president.

And in Ohio, Indiana, and Kenutcky where a combination of unpopular Republican governors and Bush fatigue are are driving voters away from the REpublican party.

Barry Welsh in the IN-6 ran a poll, and he found with most of the sample in that he was at 26% his opponent at 38%, and undecideds at 22%. Barry still hasn't released the full results, but this is a a district that the Rep. won 67-32 in 2004.

I think that we are going to see a wave to undo 1994 this year. I think that the voters who went for Perot in 1992, then turned to Republicans in 1994, are going to either sit out or voter Democrat this year. And I think that these surge voters who switch parties in watershed years like 1974 or 1994, then stay the course until another watershed year will decide the election. And I think that they're concentrated in places like Indiana that are "red" states not "blue" states when we talk about Presidential vote.

When Indiana redisticted in 2001, the guy drawing the maps used a average of 3 statewide low profile races (Treasurer, Clerk of Courts, etc.) to attain a measure of partisanship, he choose these because those are the races where party is the determining factor. I just don't understand people who think that Presidential vote is an accurate measure of underlying partisianship for Congressional Districts. It's the worst possible measure.

by ManfromMiddletown 2006-09-21 05:45PM | 0 recs

permalink

Re: Surge years

The presidential vote is hardly the worst possible measure of partisanship. The percentages in a gov race would be exponentially worse. Just think if someone who knew nothing about American politics tried to evaluate each state or district right now via recent gov results. Red Massachusetts and blue Oklahoma, etc.

I agree the lesser statewide races can be a good measure. In '96 I looked at many factors trying to get an advantage in an election pool. The problem with lesser statewide races was you occasionally had a young outstanding candidate who was using that race as a stepping stone and it would really throw off the percentages if he won big. And that would happen quite often, especially if he was a son or relative of a bigwig in the state's recent political history.

I ended up using the presidential vote to evaluate states and districts and I've never regretted it, especially in that 16-man election pool.

by jagakid 2006-09-21 07:18PM | 0 recs

permalink

Re: Surge years

You miss the point.

The reason why you don't use the Presidential vote totals, the governor vote totals or any other high profile race is because these totals don't reflect how person would vote absent information.

In Presidential and governor races the costs to the voter of gathering information are extremely low, so they make a vote based upon their assessment of the candidates. In low profile races, they have no information they vote based upon the way that they feel about the parties. Once you know how voters will vote in the absence of any other information but party, you have the partisan makeup. You can't get that from presidential vote totals.

The problem is that if you use the presidential vote total you write off districts like the IN-8 and IN-9 where Kerry lost by 16%, but actually lean democratic in their partisan makeup. Our gains this year are going to be in heavy red seats that Kerry lost by wide margins, and I woulnd't be surprised if we lose seats in blue areas. And it that happens, what use is it to call an area blue because it voted for Kerry, when they send a Republican to Congress?

by ManfromMiddletown 2006-09-21 07:40PM | 0 recs

permalink

Re: "Macro" Factors

Chris,

It wasn't an insult. Jez, chill.

I was referring to the fact that every election cycle we get bombarded with statistics that in the end don't mean very much.

Lies, damn lies and statistics. Understand?

I especially hate when we get these college professors with their election "forecasts." I mean have they EVER been right?

SO i'm sticking with the fundamentals. What are the polls showing? Who supports who and why, etc?
Are the undecideds Democrats or Republicans? If Republicans, are they undecided because they are just holding back their support or is there something going on within the Republican party? If they are independents is their a shifting wind which may prove a huge turning point across the nation? Why are they supporting us/not supporting us?

That's all.

by JackBourassa 2006-09-21 05:03PM | 0 recs

permalink

What on earth are you smoking?

There is a clear correlation between presidential approval and midterm success. There are some outliers, sure, but there is a pretty obvious correlation.

No offense, but this post is pretty much a waste of time. You grab some data (your chart doesn't even include electoral success, just approval. I'm not sure what it's supposed to even show) and then pick a few examples based on the last two cycles.

Historical data shows a clear correlation. here is another link. All I did was google for "presidential approval correlation midterm elections" to find that.

Seriously.

by delmoi 2006-09-21 05:16PM | 0 recs

permalink

Re: What on earth are you smoking?

No, there is no clear correlation. Look at your own graph. There is a vague rule, but it doesn't even come close to corrleation. You only have 17 data points. Six of them are huge outliers. That isn't a staistical correlation.

by Chris Bowers 2006-09-21 05:25PM | 0 recs

permalink

Erm

Most of the outliers are very old. Of all 17 datapoints, the correlation is 0.47. Using just the past 10 elections in the dataset, the correlation is 0.79. I would hardly call that "not there".

by delmoi 2006-09-21 05:46PM | 0 recs

permalink

Re: Erm

A dataset with only 10, or 17, reference points can never be "there." 0.79 with only ten points is not something I am willing to go out on a limb with when it comes to a forecast. Hell, even if it was 1.0, I doubt I would go with it if there were only ten points.

What would the graph look like if House results from non-midterm elections were included?

by Chris Bowers 2006-09-21 05:56PM | 0 recs

permalink

Re: Erm

I'm not saying you should, but statistics is a branch of mathematics. You can't just eyeball graphs and say whether or not there is a correlation.

by delmoi 2006-09-21 06:40PM | 0 recs

permalink

Re: What on earth are you smoking?

yeah, Chris, you should really read up on your statistics. maybe what you mean by "clear" correlation is really "deterministic" correlation. i think we can all agree that multiple factors influence electoral outcomes. the question of what determines presidential approval, or congressional returns, is complicated. you can't just dismiss a factor like that just by looking at graphs.

for example, its mathematically possible for two variables to share all sources of variance, and still look only weakly correlated when plotted against each other. this is especially possible if there are many sources.

the issue of "macro" factors really gets at things like timescale and "prior" expectations. district-by-district polls only tell you how things are currently. they do not capture, for example, inherent advantages Repubs have had with cash, which really makes a difference in the last weeks and definitely will this cycle. another example: stable advantages GOP has had on terrorism, and the chances that some "events", manufactured or real, would give them an opportunity to beat on it even more.

so if the question is "who will win in November", or "what's the current trajectory for Dems", you need to take into account much, much more than polling data. you also need much more than presidential approval, since that also changes. just witness the almost-continual rise since early May.

statistics don't lie if you use them correctly and know their limits. but that seems rare, especially in the press, which seems eager to read tea leaves.

by Chris G 2006-09-21 07:14PM | 0 recs

permalink

What always is missed

in these discussions about job approval and mid-terms is they forget a key variable: how many seats did the party in power pick up in the preceeding election?

2002 may be the single most mis-understood election in history because people forget that the GOP picked up only a handfull of seats in 2000.

The intuition is as follows: when a party wins the whitehouse coatails cause that party to pick up seats in marginal districts. In the following mid-term the party in power loses a portion of these seats.

by fladem 2006-09-22 10:17AM | 0 recs

permalink

Re: "Macro" Factors and Forecasts

One factor that is different is that the Republicans control the Presidency, the Senate and the House in '06. The big '94 Democratic loss came when the Dems controlled everything and then BOOM - we went from Speaker Tip O'Neil and Senate Leader George Mitchell to Speaker Gingrich and Leader Trent Lott overnight.

Reagan had a Democratic House his entire Presidency. At a glance it looks like divided government moderates the correlation between unpopular Presidents and off year House gains. One party rule may tie the House more to the Presidency.

by joejoejoe 2006-09-21 05:30PM | 0 recs

permalink

"Macro" Factors and Election Forecasts

Hm, I'm not so sure about this one, Chris. I don't think you should be looking at how many seats total were won during each election, but rather the net change in seats. You cite 2002 vs 2004, 1994 vs 1998, and 1982 vs 1986 as examples, so let's take a look:

2002 vs 2004

For example, in 2002, Republicans gained 8 seats with a high Bush approval rating. In 2004, if you were to subtract the Texas stuff (since it's not indicative of any national trends, just regional screwiness), Republicans would have lost seats that year, with a <50 Bush approval rating.

1994 vs 1992

In 1994, Democrats lost over 50 seats. In 1998, Dems gained 5 seats.

1982 vs 1986

In 1982, with a low Reagan approval rating, Dems gained 27 seats. In 1986, with a high Reagan approval rating, Dems gained 5 seats. This last one is a bit of an outlier since it would have made sense for Dems to lose some seats with a high Reagan approval rating.

I'm not saying Presidential approval rating is entirely indicative of how many seats will be won, but if you look at it in terms of net loss/gains, there's much more of a correlation than what you've shown above.

by Fran for Dean 2006-09-21 05:58PM | 0 recs

permalink

Damn, You're Right

Why didn't those older generation analysts ever think of looking at it district by district. You young folk come up with the most innovative things!

by kaleidescope 2006-09-21 06:39PM | 0 recs

permalink

Election Forecasts

Chris,

It might be interesting to take a look at the various races from an ideological perspective. For instance, replacing Lieberman with Lamont moves us closer to a Senate with progressive values. I'm hoping that even if we find ourselves disappointed with the balance of power in terms of raw numbers of R's vs. D's on Nov. 8, we might see an ideological shift that will represent progress for our side.

by global yokel 2006-09-21 07:57PM | 0 recs

permalink

Then again

It seems the data on this questions is very thin and noisy and thus not meaningless just not strongly correlated. However, a few observations seem in order.

1) Bush is much more strongly identified with the Republican party than Clinton with the Democrats. Not that Bill wasn't a Dem, its just that he spent a lot of time on NAFTA and Welfare Reform etc.

2) The reason people dislike Bush - primarily the war and Katrina - are also strongly associated with the Republican party. So his dislike is representative of a general dislike of the Congress and republicans.

3) Really, though, lets face it. Bush in the high thirties, low forties is a dream come true - or a nightmare for the country.

by Wesc 2006-09-21 08:57PM | 0 recs

permalink

Question for you experts

Guys, I'm not an expert; I just follow politics. So I have a question: Doesn't the fact that the Democrats rolled over for Bush for most of his two terms contribute to the fact that Bush's low ratings may not appear to affect congressional races? If the Dems had actually been a legitimate opposition party, wouldn't they have been better at linking Bush with Republicans in Congress?

Bush's ratings are at best where they were before 9/11 and many polls have him mired in the 30s. Jesus, you're telling me that a president this unpopular doesn't drag down his party in elections? If not, it's because of the trifecta of modern politics: the Democrats are gutless and incompetent, the White House gets away with lies and propaganda, and the media abets the GOP lies and propaganda.

Without looking at other factors, Chris's numbers don't seem to make much sense to me. I'm 58, and other than Nixon, I've never seen a president this unpopular -- even despised by many people. And I've never seen a president whose policies were such abject failures and who tried to undo policies and programs that are basic to America. If this stuff isn't having an effect on congressional elections, there is a reason. Maybe I've lived a delusion all my life, but when presidents fail at this level, it usually hurts the politicians in his party.

I'm not saying Chris won't be right. I fear he will be. But I guess I'm saying it won't be because it just is, like gravity, a law. It's because of other reasons such as the ones I've enumerated. The bottom line is that we're in danger of becoming a de facto one-party country.

I'd appreciate somebody helping me to get my mind around this issue.

by Phil from New York 2006-09-22 07:42AM | 0 recs