The “Wall Street” effect on incumbent electoral success in the US House of Representatives

Before returning to graduate school to obtain my Master of Public Policy degree at Georgetown, I spent several years working in the financial services sector, where I researched the financial markets in general, and the US equity markets in particular.  I saw firsthand the effects the financial crisis of 2008 and 2009 had on investors and their portfolios.  Much of my studies have involved how Wall Street and the federal government interact, especially with respect to the Federal Reserve and the Treasury Department.  I became interested later in how legislators are affected by Wall Street, but not in the ways we usually think about: campaign contributions and lobbying.  What I want to see is whether the electoral fortunes of representatives are actually influenced by how well the stock market actually performs in an election year.  Do voters, on average, vote their portfolios as well as their pocketbooks, holding other factors constant?

Much research has been done regarding the effects of macroeconomic conditions on the electoral chances of congressional candidates, in particular incumbents, with mixed results.  However, published research thus far has ignored effects of equity market performance on congressional elections.  As more and more Americans have become members of the “investor class,” it stands to reason that voters might punish incumbent politicians, in particular those of the current president’s party, for how well their portfolios have performed just as they do when the economy weakens.  In other words, are do we ascribe too much of the probability of reelection of incumbent congresspersons to the economy because we don’t account for voter reaction to the stock market?

The question of whether or not there is a “Wall Street effect” of market performance on the electoral chances of incumbent House candidates is substantively important on a few levels.  If we assume that incumbents are self-interested with respect to their own reelections, being able to identify effects resulting from stock market performance could lead to incentivizing policies that promote weak market performance by out-party members.  Given the trend of divided government with the House of Representatives being controlled by a different party than the president, these effects could lead to out-party incumbent members supporting policies that may in fact be detrimental to market performance which could damage in-party electoral chances while helping their own.  In sum, this effect, if it exists, could create perverse incentives for a large group of powerful politicians to encourage weak performing markets.  From a policy perspective, it stands to reason that political strategists both on Capitol Hill and in the White House would tend to want to know how to frame the debate with respect to markets in election years.  Furthermore, if these effects are shown to exist, future congresses and presidents might reassess their own election tactics in light of available information.

I hope the results of this research add to a colorful and necessary discussion regarding Wall Street and congressional self-interest.  I am also looking forward to hearing about the other fellows’ research interests and hope to provide any guidance I can.


About jimsteiner

I am currently a graduate student at Georgetown University studying Public Policy (MPP 2012) and concentrating in public management and leadership. I have an undergraduate degree in business administration/finance from George Washington University (c/o 2003) and spent seven years in the financial services industry. I spent this summer studying international management at Corpus Christi College, University of Oxford. My PFP research proposal involves empirically examining effects of stock market performance on incumbent congressional reelection rates controlling for macroeconomic and political environments. Upon graduation, I intend to seek a career in strategic management consulting or in management in the federal government.
This entry was posted in Economic and Financial Policy. Bookmark the permalink.

6 Responses to The “Wall Street” effect on incumbent electoral success in the US House of Representatives

  1. Jonathan Robinson says:

    Interesting proposal, though as you say there’s a lot of work done on pocketbook voting…and how there are limited effects. Though a paper you might be interested in is this new one by Bartels and Page on the preferences of wealthy Americans (the new investor class)

    I’d be interested in seeing how the supposed rise of the investor class (if at all) plays out politically. Most Americans aren’t in the stock market, and tend to not pay attention to changing performance.

    Final thought. Why would we expect equity markets to be different from an index?

  2. jimsteiner says:

    I’d take issue with your assertion that most American aren’t in the stock market. A majority of Americans have been invested in stocks since at least the late 1990s (peaking at 67% in 2002). It’s dropped to about 58% in 2011 as of April. Interestingly, this is roughly the same figure for national turnout in the 2008 presidential election and far higher than midterm election turnouts. Because it’s there is a likely correlation between equity participation rates and turnout, I intend to control for both in my regression model (along with several other factors, like market fear, war, currency strength, party strength, split government, and so on).

    Whether most Americans tend not to pay attention to changing performance goes to my research question is a way. If I can identify a statistically significant effect of market performance on incumbent reelection rates, ceteris paribus, then I could make the argument that, contrary to popular belief, people do pay attention to performance of the stock market. I have another hypothesis that ownership rates may not even matter. Because people in general are not given the appropriate economic data to make informed decisions, even when they say the economy is the number one issue in an election year, they will use readily available proxy for this information like stock market performance. I intend to control for volatility as well to see what is driving these changes.

    I have yet to run regressions yet since I am still building my data sets, but it will be interesting to see what comes out, either way.

    Finally, I am not 100% sure I understand your question regarding equity markets versus an index. I don’t believe I made any such distinction. The S&P 500 is an index that I could use as a good proxy for “equity markets” since it is highly correlated with the population of all US stocks and has little tracking error. So I would say that I am emphatically NOT making a distinction. In fact, the converse case is true. If I am not interpreting your question appropriately, please let me know!

    Thanks for taking the time to reply. I find it very helpful to think about my premises from other perspectives as I am developing my research.


  3. Jonathan Robinson says:

    1. So I’d take the fact that stock ownership is similar to turnout probably as spurious. If you look at this website, you’ll see that stock and equity ownership is concentrated amongst the wealthy (the demographic of your average equity market participant is probably a bit different from your average voter).

    2. To the best of my knowledge ownership in equities is an artifact of employee mutual funds and retirement defined contribution plans. Just reading the Executive Summary of this study, revealed the story is more complex (look also at pg 25 of the PDF)

    3. Bah! In regards to my last point I meant, why would we expect equities to perform differently from an index of macroeconomic factors. It would seem to me that the S&P would pick up lots of white noise and is hard to explain. I would also guess that the stock market correlates pretty well with macroeconomic factors too…I guess I”m just suspect that the stock market is really consequential.

    Just playing devils advocate. I’d love to be proved wrong.

  4. jimsteiner says:

    Two things: first, playing devil’s advocate is what the peer review process is all about, not to mention it helps to focus one’s research question and revisit assumptions and premises. I’ll have to defend my thesis eventually, so it helps to do this every step of the way. Two, there aren’t really any statistical tools available to prove you wrong (or right). If there were, I’d not waste a whole lot of time bothering with building data sets! All I can do is attempt to determine whether there are statistically significant relationships between variables after controlling for other effects. I hope to find statistical significance, which would support my hypothesis, but even not finding stat sig effects is also important in order to look for other effects. The advantage of doing several regression models is that I might find stat sig effects for control variables which may explain even more variation in my dependent variable than I hypothesize.

    To address your points:
    1. I was not intending to make any causal or even associative claims between stock ownership and voter turnout. I must, however, include both in a model if I think there is any chance they could be correlated or risk biasing my estimators. Since both proportions are above 50% it’s probable that there is at least a small correlation, though I won’t know for sure until I run a correlation test and test for significance.

    2. I agree with that claim. It shouldn’t make a difference if we think people care about the performance of their 401k just as they would a brokerage account. Anecdotally, since you can deduct losses in non-retirement accounts, I’d assert people might care MORE about deferred retirement assets. Furthermore, the effects might even be more pronounced at mid-range income levels than high income levels since the wealthy are more likely to be able to adjust to market movements (the nature of 401ks vs IRAs is that IRAs can buy and sell at will, whereas mutual funds in 401ks are not technically traded and only redeem at the end of a market day).
    3. Equity markets perform “differently” from macroeconomic indices in the sense that a soft economy does not necessarily imply poor performing markets and vice versa, but that is not to say there is not a correlation between such indices and the equities markets. Because equity markets are considered to be leading economic indicators, in fact, you will see a statistically significant correlation. This is the whole point of controlling for macroeconomic indicators in the first place. That is, I want to see if we hold all economic factors we think are correlated with both the likelihood of reelection and equity market performance constant (along with other independent variables), we can isolate the portion of the variation in the probability of reelection that belongs exclusively to market performance and nothing else. Any observed effects may or may not be statistically or substantively significant (which is the actual point of doing the regressions). In any event, all of this goes to the heart of the question, which is whether voters use the markets as a proxy for macroeconomic conditions since they are easier to interpret (though not necessarily accurately)—regardless of personal equity stake—when they decide for whom they will vote. In order to observe effects that are unbiased, one must necessarily control for those things we think are correlated with both the probability of reelection and with market performance.

    I am not so sure I understand what you mean by “white noise” and how the S&P is hard to explain, so if you want to clarify, I can try to address those points.


  5. Jonathan Robinson says:

    White noise, ie…markets react in a herd like manner. Many times they fluctuate based on unexplainable factors that have little to do with the macroeconomy. Sure they exhibit trends but the fluctuations of the data are far more severe than with BLS data on unemployment for example.

    A sidenote but something to remember whenever doing research, causation is as important. It’s something I write on a post it note whenever I’m doing research.

    • jimsteiner says:

      For my hypothesis, the source of variation in equity market performance is irrelevant. I am only attempting to isolate whether or not voters punish incumbents for poor performance. It doesn’t matter necessarily whether it’s from the economy or a terrorist attack or high oil prices. The “white noise,” as you put it is what we call “unexplained variation” in econometrics and is actually the key to answering our questions.

      For the sake of argument, let’s consider a simple model. Say my dependent variable is “probability of incumbent reelection.” There are myriad factors that are correlated with this variable (in other words, explain the observed variation). These could be things like party strength, divided government, or other political factors. My independent variable of interest here is equity performance. If I were to run a simple bivariate regression on “reelection” and “equity performance,” I would see an effect (a coefficient). However, this would be biased (in other words, the observed effect is not the “true” effect), unless we have controls for things that are correlated with “reelection” and with “equity performance”. The idea is to identify and include in your model all of those variables. When we control for these variables (if we first identify a correlation between the IV and DV), we let the proper amounts of variation we see in the DV be ascribed to the proper IV’s.

      To illustrate, say I ran that simple bivariate regression and it shows that each percentage point decrease in the market predicts a decrease in the probability of reelection by one percentage point. This would be interesting, but the observed effect is likely different from the “true” effect in the population. If I think macroeconomic factors are correlated with both things and I control for it, I might now see that the new variable accounts for, say, 40% of the variation. And let’s all other variables which are correlated with both reelection and equity performance are in our model (not possible in real life, but this is theoretical) and together, all these account for, say, 20% of the variation. So now 60% of the variation observed on our dependent variable is accounted for and there are no other variables out there that are correlated with BOTH the independent and dependent variable. This would mean that whatever effects I’d see of market performance would be “unbiased,” that is, the effect we’d see after controlling for other factors would be the “true” effect that belongs only to market performance and no other variable.

      Now, we could certainly include variables that are correlated with reelection but NOT market performance (remember, we already controlled for all those), and this would make the model more efficient and would explain more of the variation on the dependent variable. However, none of these variables would change the effect of market performance (since they are not correlated with it).

      Finally, I am not attempting to identify or even make a claim for causation. I am interested in whether there are statistically significant effects observed. And, as you alluded to in a previous post, there could be spurious effects which cause both market performance and reelection rates to change. It is extremely dangerous to make causal CLAIMS (versus statistical inferences) in social sciences, unless you are running randomized controlled trials, which are impossible for this data.

      In any event, it’s generally inadvisable at this point to discuss causation outside the context of your research question, especially without observing the regression output, testing for statistical significance, and making inferences.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s