The Incredible Statistical Difference between Democrats and Republicans
by Benjamin Studebaker
Lately I’ve been reading a book by Larry Bartels entitled Unequal Democracy: The Politics of the New Gilded Age. Bartels has conducted an incredible study that produced statistical outcomes I was not, to this point, aware of. Given the amount of time and energy I devote to politics, it stands to reason that the general public is not aware of them either, and it is absolutely imperative that, so long as this remains a democracy, every citizen be made aware of what Bartels has found, so I set out today to communicate these figures to whoever might be out there reading, in the hope that perhaps they will be shared more widely.
Here’s what Bartels did–he took US economic data from 1948-2005 and separated it out by which political party controlled the white house. Truman, Kennedy, Johnson, Carter, Clinton go to the democrats, Eisenhower, Nixon, Ford, Reagan, and the two Bushes go to the republicans. He compared the economic numbers under the democratic presidents to the numbers under the republican presidents. He included a one-year lag under the assumption that policies take some time to go into effect–the first year of Reagan’s presidency counts as part of Carter’s, the first of Carter’s is Ford’s, and so on. Bartels went into this study with non-partisan scientific intentions. He claims to have not voted since 1984, when he voted for Reagan. Nevertheless, he found stunning differences, far too strong statistically to be driven by mere chance–it held true even if any one or two presidents were excluded from the study.
A few of Bartels’ graphs have been posted elsewhere online in years past, but some are nowhere to be found, so I have reconstructed them on excel. This first figure compares real (inflation-adjusted) income growth for several different income percentiles (the 20th, 40th, 60th, 80th, and 95th) under presidents of the two parties:
This reveals two spectacular differences between the parties, one of which most of us likely suspected but could not prove for sure, the other likely violates the expectations even of many on the American left:
- Under democratic presidents, economic inequality decreases slightly. Under republican presidents, economic inequality rises rapidly.
- Democratic presidents produce higher average annual income growth for all portions of the population than do republicans–even the 95th percentile experiences higher income growth under democrats.
It’s possible that there is some portion of the population even more affluent than the 95th percentile (the top 1%, 0.1%, or .01%?) that experiences higher income growth under republicans, but this nonetheless suggests that even someone in the top 5%, with a personal income of $100,000 or household income of about $170,000, will prosper on average slightly more under a democrat than he would under a republican. If citizens voted for the party that put more money into their pockets, at least 95% of them would vote democratic every time.
Many commentators believe that rising inequality is structural, that certain rare or difficult to acquire skills are currently prized by the economy, and that this is the reason inequality has risen post-1980. Bartels’ numbers directly challenge this claim–if inequality were structural, we would expect to see it grow at roughly the same rate regardless of party politics. Instead, we see a vast difference between the two parties–republicans expand inequality at breakneck pace and democrats shrink it slightly.
In his book, Bartels goes to great lengths to verify his result. He adjusts for a bevy of possible noise-making variables–unemployment, inflation, oil prices, labor force participation, and so on. If readers doubt the veracity of the figures, I encourage them to read his book in which he discusses his methodology in much greater detail than I can here.
Republicans could respond to these numbers by arguing that their presidents are forced to embrace contractionary policy in order to combat the runaway inflation they believe democrats permit. However, Bartels does not merely compare incomes–he also looks at the difference in the unemployment rate, inflation, and GNP growth, and those results are just as stunning:
This reveals that while democrats keep unemployment rates significantly lower than do republicans and maintain much higher average GNP growth rates, the republican advantage in inflation is statistically insubstantial. This suggests that there is no substantive economic advantage to having a republican president at all.
However, despite this, democrats won presidential elections only 40% of the time during the 1948-2005 period (they went 6-9, in sports terms). Why might this be? Bartels observes that republicans regularly generate much higher income growth rates during election years than they produce the rest of the time:
Amazingly, in a non-election year, an American in the 20th percentile actually sees his income shrink on average under a republican president, yet in an election year, his income will grow as much as the income of someone in the 95th percentile does in an ordinary year.
By contrast, democrats have consistently failed to manipulate the voters in this way, actually performing worse, across the board, in election years:
This is very odd. We know democrats know how to grow the economy more effectively than republicans–they get better numbers most of the time. Yet, during an election year, republicans can spontaneously turn it on while democrats run into some kind of obstacle. There are a few possibilities:
- Ignorance–republicans are bumblers, but they have the support of some plutocracy (the top 1%, 0.1%, or 0.01%?) that is in charge of a sufficiently large part of the American economy such that they can manipulate wage growth figures to make republicans look good during election years and make democrats look bad, especially to potential affluent donors who actually experience a decline in real incomes during democratic election years.
- Malevolence–the republicans deliberately grow the economy slowly for some reason (perhaps inflation? But we’ve already established that they don’t fight inflation effectively). In election years, they deliberately change policies, producing higher growth figures temporarily in order to win. When in opposition, they use congressional influence to sabotage the democrats’ chances.
In either case, the American public is being played–the question is by whom (the very rich, or republican politicians?). If republican politicians are doing this, it unveils a shocking asymmetry–democrats either do not engage in this tactic at all, or do so with spectacular incompetence. I think it is more likely that republican politicians are bumblers aided by the machinations of super-rich backers, though it’s possible that both forces are to some degree in play.
Given this information, it appears masochistic to support republican presidential campaigns, yet inevitably, large portions of the population continue to be taken in by the illusions put in front of them during election years. So long as the United States remains a democracy, it is imperative that as many of them be made aware of the fact that they are being manipulated in this way as possible. If you agree, please share this piece.
UPDATE:
I thought about this a little more and decided that the two explanations for the gap in election year performance between the parties I offered here were flawed. I have written a more detailed explanation that is more attentive to the contingent qualities of the various elections in Bartels’ data set. It can be read here.
Benjamin Studebaker 
I have not read Bartel’s book and so my comment is geared towards your analysis in this blog than towards the original content.
Presidents have limited control over the economy. Congress has much more of a direct impact, but even they have limited direct control. Government’s role in the economy is substantial, but not the only factor at play by any means. Although the correlation presented here between party of the president and economic growth may be strong, this is far from indicating causation. It would be tempting to some to suggest that Gingrich’s Contract with America might have had more to do with economic growth during the CLinton years than Clinton, for example. Or we might just say that investments in science and technology pay off. The statistical story being told in your post is interesting, but its relevance is questionable.
I certainly agree that presidential power waxes and wanes and is not at all the only or definitive influential factor. Nonetheless, the sheer strength and consistency of the statistical relationship in Bartels’ data makes it very hard to deny. Bartels calculates that the probability of a correlation of this strength arising through random factors disconnected to presidential policy to be about 0.006. If you find my defense unpersuasive, I would encourage you to read the original text, linked in the piece.
Correlation is not causation. In my opinion you read far too much into these statistics without further analysis of variables with much more material power to influence the economy. Dem v. Rep is far to broad a category to suggests manipulation or incompetence.
This is a classic problem with political thinking in the realm of economics – suggesting politicians control an economy when much larger shocks are happening. You fail to even note monetary policy which aside from the appointment of the Fed’s head, the president has zero control over.
Your comment suggests an r^2 of 0.994 which adds to my skepticism. Anyone who has preformed a regression knows how high a number that is and how improbable it is. Lines like the one below make me scream inside. You’re dealing with a very small sample and excluding sooooo many variables.
“This is very odd. We know democrats know how to grow the economy more effectively than republicans–they get better numbers most of the time.”
There are of course significant difficulties with performing any statistical inquiry in political science because the time scale under which the variables and conditions are even slightly consistent is very short (looking at presidents pre-WWII presents a vast array of problems). That said, the research Bartels has conducted is the most methodologically rigorous research of this kind possible in this field of inquiry.
I encourage you to have a look at Bartels’ book. He controls and adjusts for many variables and considers a wide array of possible alternative causes. It is not realistic nor is it useful to hold empirical statistical research in political science to the same standard one would hold it to in the natural sciences. The real world is not a laboratory.
John Smith is quite correct that a general failure to consider mechanisms in economic analysis is a real problem, and I would beg to differ that it is not possible to hold economics and social analysis to the same standards as the natural sciences Indeed, I would argue that our current difficulties arise from our collective failure to do just that. From the details in this posting though, I would certainly not dismiss Bartels’ claim.
What are actual mechanisms that could be involved in this? Growth in income, and growth generally in any measurement made in money, is a function of the money supply, and distribution of that money across the economy. Growth in the money supply comes primarily from commercial bank operations, and in particular lending. For the USA it’s typically about 6-8% per annum, and that figure is remarkably stable across time – so we can’t accuse the different parties of messing around with the money supply, as does happen in some other countries. However, it would be a mistake to assume that that creation is evenly distributed across the economy, a lot of the ‘wealth’ of the top percentile is in fact being created by lending.
What does differentiate the two parties is taxation and government subsidies. If we think of all the money flowing around in the economy as a fluid, very generally the democrats tend to take a proportion from the people with the most, and feed it back into the general economy via government spending, republicans tend to favour policies that don’t do that. I don’t it’s too hard to visualise what’s going to happen over time, as money is created, and funnelled through borrowers into a sub-section of the general economy, and there is no counter-process to redistribute it more generally.
I’m not sure how Bartels could get the sample size necessary to do this study in a way that would sync up with natural science standards. The .006 figure comes from there being only 11 US presidents during the postwar period in 2005, the last year of data Bartels was able to run. By the time there would be a large enough sample, not only would the surrounding economic landscape be very different, but all of the people who are currently affected by policy would likely be dead.
I agree that fiscal policy is likely the primary mechanism by which US presidents exercise influence (the progressiveness of the tax code, the distribution of spending in the budget, stimulus/austerity preference, etc.). However, I do think there is some scope for monetary influence as well–presidents tend to appoint Fed chairmen who match them in hawk/dove mentality. Bartels points out that most of the US recessions during this period resulted from deliberate interest rate hikes by the Fed to curb inflation.
The three pillars of science are Theory, Observation and Experiment. As you point out, Economics historically has had to rely on two – Theory and Observation. The problem there is that a lot of different theories can be made which all fit the same observation, and here we are. However, with a realistic simulation of the monetary system (obligatory disclaimer – this is something I’m working on at the moment), then it’s pretty straightforward to setup an experiment to show effects like this, and to rule out (or in) different possibilities.
It turns out it’s even more straightforward to setup experiments disproving some popular economic shibboleths, notably that hiking interest rates actually has no effect on the supply of loans (and hence the supply of money) from the banking system, and indeed that there may even be second order effects that cause the system to react in precisely the opposite way, i.e. raising interest rates increases the monetary expansion rate.
It’s a very slow system (also very easy to show with simulation), most of the interesting behaviour is linked to loan duration – so the one thing that can’t be ruled out is a long-lagged reaction-counter reaction cycle being inadvertently triggered by the parties policies – I’m afraid I don’t know the historical distribution of the american parties control of the various parts of the US executive well enough to comment on that.
[…] have been somewhat dissatisfied with the explanations I offered a few days ago for the surprising gap between democratic and republican economic performance in […]
[…] The Incredible Statistical Difference between Democrats and Republicans (Dec 28, 2013, 1,101 hits) […]