Yet again and again [economists and other researchers not named Hacker or Pierson] have found themselves at dead ends or have missed crucial evidence. After countless arrests and interrogations, the demise of broad-based prosperity remains a frustratingly open case, unresolved even as the list of victims grows longer.
All this, we are convinced, is because a crucial suspect has largely escaped careful scrutiny: American politics.
– Jacob Hacker and Paul Pierson, Winner Take All Politics
Here's a chart showing trends in the share of income received by the top one percent for all the modern industrialized nations for which data is available going back to the early twentieth century:
The data is from a new website created by several of the leading scholars studying inequality with tax data. The American trend, the thick black line, is from the much cited work of Thomas Piketty and Emmanuel Saez, which is part of this new database.
From 1910 to 1970, American inequality trends follow the broad international pattern, and inequality levels are in the middle of the pack. That's basically still true from 1970 to 1986:
It's rising a bit over the period, but only by a percentage point. Note I'm keeping the scale of the charts the same for each one. Here's the chart for 1988 to 2006:
Uh-oh. Now we look like our inequality levels are higher than everywhere else. What happened? 1986 to 1988 happened, as is evident from the 1970-2006 trend:
Wow, that's a four percentage point increase in two years—three times the increase over the 16 years from 1970 to 1986, and bigger than the 12-year increase from 1988 to 2000. Huh. There are two possibilities here. One is that the data is right. You can see where I'm going here.
It helps to know that the 1986 tax reform created big incentives for people who had previously reported income on corporate returns (where it is invisible to the datasets above) to report on individual income tax returns (where it appears as an out-of-the-blue increase). And if this may be considered a permanent change in the tax regime, then the effect is for more income to show up on individual returns after 1986 than before, artificially lifting the top income share in every subsequent year.
Hmmm...which possibility is more likely? Let's look at another chart showing the trends just for the northern hemisphere Anglophone countries, to which I'll add a new line:
OK, from about 1940 to 1986, these trends line up strikingly, then the U.S. trend goes AWOL. However, let's instead assume the post-1986 U.S. trend is an artifact of the 1986 tax reform. First, let's increase the top one percent share from 1986 to 1988 by the same rate that it increased in the U.K. Then let's let the top share in the U.S. increase by the same rate that it actually did from 1988 to 2006, but from the new, lower 1988 level. The result is the revised line above. This makes the U.S. trend and level consistent with not just the U.K., but Canada.
Of course, if the 1988 to 2006 top share levels are more accurate in the U.S. after 1988 than before 1986, then rather than lowering the post-1986 trend, we should raise the pre-1988 trend. That would make U.S. levels uniformly higher than in the U.K. and Canada. But of course, the measured U.K. and Canadian top share levels may also be artificially low due to tax avoidance. And of course, the common trend over the three countries would remain.
So, to review, when the post-1986 U.S. trend is corrected, the U.S. experience with inequality over the past 100 years is broadly consistent with the rest of the modern world. Here's the summary chart for 1910-2006, with the revised U.S. trend.
Comparing levels is more difficult, but many recent cross-national comparisons related to inequality are about why trends differ. What these five charts clarify is that explanations for the recent rise in American inequality that focus on uniquely American causes—such as greater political muscle-flexing among corporations and the mega-rich—are insufficient (and unnecessary).
Update: I've received several responses offline that it's going to far to say the experience of the U.S. is like that "everywhere else" and that it is really only like the other Anglophone countries. To some extent, that's a fair criticism. But of 15 countries shown here, only Germany, the Netherlands, and Switzerland haven't experienced an increase in inequality since 1980. And the increases in Norway and Finland are as big or bigger than in the U.S., U.K., and Canada. Sweden's increase is also nearly as great in relative terms (starting from a much lower level of course). But even if this is a story about the U.S., U.K., and Canada or the Anglophone countries versus the rest of the world, that's still a problem for Hacker's and Pierson's U.S.-centric theory.
My continually interrupted consideration of the new Jacob Hacker/Paul Pierson book Winner-Take-All Politics has led me to go back to the original inequality estimates of Emmanuel Saez and Thomas Piketty to understand better what they did. And what they did involves making a lot of assumptions about the size of their baseline populations (from which the top 1 percent is identified) and about the size of total income in each year (to which the top 1 percent's income is compared in order to get the top 1 percent's share of income).
I get really nervous about these kinds of assumptions, so I went looking for some alternative data sources. I remembered that the Federal Reserve Board's Survey of Consumer Finances explicitly samples very rich families separately from its primary survey efforts, drawing names of rich folk from IRS data and then tracking them down to interview them. This data is not perfect either--the response rates among the very rich are very low, and comparable data only go back to 1988--but the SCF has a number of features that are preferable to the Piketty-Saez dataset. For one, it provides direct estimates of the baseline population's size and of the population's total income. No need for worrisome assumptions to compute the top-share estimates. Second, a more comprehensive income measure may be used, including public transfer income, such as unemployment insurance benefits and worker's compensation, and retirement income from Social Security and drawn down from private pensions.
The Fed's main SCF analyst, Arthur Kennickell, put out a paper last year that provides top income shares for 1988 to 2006 (see Table 4). I plotted those against part of the Piketty-Saez series that includes realized capital gains, and here's what I found:
Awfully consistent....What this tells me is that the concern I expressed in my Hacker-mania post that the Piketty-Saez series inflates the top income groups is probably not that big a deal. On the other hand, since the SCF data only go back to 1988, changes in the extent to which the P-S series inflate the top income groups may still create the exaggeration of the increase in inequality since the 1970s. But certainly the SCF estimates should be viewed as making that less likely.So do I now think the Piketty-Saez estimates are correct? No--there is still the potentially very important issue of how people receive compensation and report their incomes on tax returns in response to changes in tax law. And that issue still affects the SCF data too because it affects what gets reported to SCF surveyors when they ask about specific types of personal income.Just as importantly (maybe more), is the fact that neither data series accounts for employer and employee contributions to benefits like health insurance, and neither accounts for unrealized capital gains that investors accrue or the returns accruing to pension benefits pre-retirement. But I feel more confident that the assumptions Piketty and Saez make in measuring the quantity they claim to measure are solid (save that hugely important assumption that tax law changes don't meaningfully affect their series).
Today, the DC progressive/new media promotional machine launches Jacob Hacker and Paul Pierson's new inequality tract, Winner-Take-All Politics, with an event at the New America Foundation. I don't want to tell you not to buy the book or that it is likely to be wrong—I've bought it myself, but only just started it. What I do want to tell you is that since Hacker has been making grand statistics-based arguments—beginning with his and Pierson's Off Center, and continuing with his Great Risk Shift—his books have been provocatively and cogently argued, have told progressives exactly what they want to hear, and have been based on statistical evidence that I have found to be completely wrong.First, in Off Center, Hacker and Pierson argued that Republican success in the aughts invalidated the "median voter hypothesis" that argues that the parties will tend to take policy positions oriented toward the preferences of moderate voters. They claimed that in recent decades, the Republican caucus had moved steadily rightward (true) while the Democratic caucus had, if anything moved rightward too (ehhh...OK). But because the ideological distribution of the electorate hadn't changed, that meant that Republicans had somehow pulled policy "off center", which Hacker and Pierson say was accomplished through various dirty tricks and hard-knuckled tactics.What actually happened is that at the start of the 1970s, the Democratic Congress was "off center"—to the left of voters—and so the rightward shift of Congress and the Republicans reflected a move that produced a Congress more consistent with the views of voters. In other words, the median voter hypothesis explains the changes rather well.In The Great Risk Shift, Hacker argued that economic volatility had skyrocketed—more than doubling between 1974 and 2002 and nearly quadrupling between 1974 and 1994 alone. Oops—these results turned out to hinge on an arcane methodological issue that Hacker should have caught. When I uncovered this problem, Hacker was forced to revise his book for the paperback edition (no, you won't find documentation that my discovery was the reason behind the revision, but it's in my in-box archives). When I produced my own estimates of income swings, I found that they had increased over time, but rather modestly, so that if a household's typical income swings were 15 to 16 percent of their income in the early 1970s, they were probably about 17 to 18 percent in the early 2000s.I also found that, contrary to Hacker's assertions, the evidence on economic risk in other aspects of life also implied fairly modest changes in recent decades.(Incidentally, I will present evidence in a month at a Census Bureau conference that Hacker's latest effort, an "economic security index" for the Rockefeller Foundation, is also botched. Full details once I have the green light to circulate them after the conference.)I hope to blog over the next two or three weeks on the new book as I get further into it, but today, let me just provide some commentary on two central claims about what has happened to inequality.Claim #1: The share of income going to the top 1 percent increased from 8 to 18 percent from 1974 to 2007—from 9 to 24 percent including capital gains.A strong case can be made that the increase over time was just four percentage points or less, not 10 or 15. In a post I did a year ago, I showed that much of this increase could be explained by two phenomena: (1) a steady rise in tax filing as "subchapter S" corporations (with income reported on individual tax returns) instead of "subchapter C" corporations (with income not included on individual tax returns and thus missing from the IRS data Thomas Piketty and Emmanuel Saez use, whose work Hacker and Pierson rely on); and (2) a jump from 1986 to 1988 in wealthy taxpayers shifting to income and stock options reported on individual tax returns from fringe benefits and stock options not reported on individual tax returns in response to the tax cuts of 1986. These insights are not mine—they come from Cato's Alan Reynolds, who has been making the points for several years now. By adjusting the trend line for these changes (using data from Saez), I showed that the change in the top's share of income probably rose only 4 percentage points rather than 10 from 1974 to 2006 (the increase from 1974 to 2008 would be similar—Saez just released his 2008 estimates).Are these adjustments warranted? Well, doing so ends up producing estimates that match the trend found by Richard Burkhauser and his colleagues using the Current Population Survey. The adjusted estimates also raise doubts about the claim that the top income share has not been higher since 1928 (since they put the top share when capital gains are excluded lower than every year between 1928 and 1941).Furthermore, while I have not seen any research examining the question, I am pretty sure that these levels—and the increase over time—would be lower with ideal data. Piketty and Saez identify the top 1 percent in their data by estimating the number of single adults and married households in the population and using that as their baseline. From there, they simply look at the richest people in the tax returns until they have a group equal to one percent of this baseline. But as Steve Rose notes in his new book, Rebound, their baseline is almost 30 percent higher than the number of households in the U.S., primarily due to multiple tax returns in households with roommates, non-married romantic partners, and adult and teenage children. Inflating the overall baseline by nearly 30 percent means inflating the number of people in the top one percent by 30 percent too, which would not be problematic except that we can presume that essentially all of the inflation in the IRS data occurs in the bottom 90 percent.Let's talk concrete numbers to give a sense of why this is an issue. In 2007, Piketty and Saez use 150 million "tax units" as their baseline population, meaning that to look at the top one percent, they need to focus on the 1.5 million richest tax returns in the IRS data. Rather than look at households, as Rose does, let's just distinguish families and unrelated individuals from each other and look at them (a more conservative approach than looking at households, since there are fewer households). The Current Population Survey indicates that there were just 134 million of these, ten percent fewer than the number of tax units. So the top one percent of families/unrelated individuals included 1.3 million people rather than 1.5 million. To know what share of income the "top one percent" received, one should look at the 1.3 million richest tax returns, not the 1.5 million richest. By looking at the richest 1.5 million, the "true" top one percent is exaggerated by about 15 percent.To back into the more meaningful figure, we can assign incomes to 200,000 people and subtract them out from the aggregate received by the top 1.5 million. A rough way to do this is to give them the average income received by people in the top five percent of income, which in 2007 was $364,000 according to the IRS data (in 2008 inflation-adjusted dollars).One other adjustment should be made—because those 16 million additional tax returns in the IRS data represent people with relatively low incomes (think teenagers and college kids), the aggregate amount of income in the bottom 90 percent is lower than in the CPS. The difference should be added to what the IRS shows as total aggregate income (the denominator when computing income shares). Making these adjustments, the top one percent received 15.5 percent of income in 2007 rather than the 18.3 percent indicated by the Piketty/Saez results. And that doesn't include the adjustments I outlined above due to tax law changes. For 1974, the figures are 7.0 versus 8.1 percent. For the change over time, I show a change of 8.5 percentage points rather than 10.2. Combine this adjustment with the analysis I conducted around the effect of tax law changes, and the increase in the top share since 1974 gets awfully small—probably less than a four-point rise over 35 years.Claim #2: The nation has moved steadily from "Broadland"—typified by the expansion of the 1960s, when most of the income gains went to the bottom 90 percent—to "Richistan", where over half the gains go to the top one percent.These numbers are actually pretty solid and robust to shortcomings of the IRS data. However, focusing on changes in the income share is actually a pretty uninformative way of looking at things. Consider the last expansion, from 2002 to 2007. Something like 60 percent of the income gains went to the top one percent. It's only slightly an oversimplification to say that what happened was that Greenspan and Bernanke juiced the economy by keeping interest rates low, which had the unfortunate side effect of sparking all sorts of crazy in the financial sector (including pay increases almost surely out of line with the value these geniuses added to the economy). This is Raghuram Rajan talking, but I think he's completely right. When businesses failed to invest, the result was a weak recovery, small income gains for most Americans, and enormous gains to a bunch of 12-year-olds on Wall St.For the median family/unrelated individual in the bottom 90 percent, the increase in income according to the CPS was from $37,000 in 2002 to only $38,000 in 2007. Of course, health insurance costs were rising rapidly during this period, so the increase in total compensation was greater. But still, warm beer.Consider the counterfactual, however. What if the Fed hadn't goosed the economy? Or what if Congress had taxed the income of financial "wizards" until the gains to the top were much smaller? Either might have mitigated the share of gains that went to the top. But neither would have helped the bottom 90 percent much, and without the monetary stimulus (and the tax stimulus from the Bush Administration), the expansion of 2002 to 2007 might instead have been the expansion of 2004 to 2007, with a prolonged recession dragging into the mid-2000s. Of course, the bursting of the bubble in 2008 would not have happened either in that case, but it's not at all clear (to me) that the bottom 90 percent would be in a better place in this counterfactual scenario, despite having successfully limited gains to the top.On the other hand, had the expansion been broad and robust, producing solid gains for the bottom 90 percent and the twelve-year-old Wall Streeters still received 60 percent of the gains, I don't know that there would be reason to be equally frustrated as many are today. Hacker and Pierson's story is about us versus them, but it seems to me that they don't persuasively defend this view. If we can have a bigger pie, but only if we let the rich have a bigger piece of it, then the whole question gets a lot more complicated. Research by my former advisor, Christopher Jencks, indicates that higher inequality doesn't seem to increase a country's growth, but nor does it hurt it. I'm with Dalton Conley—we should care less about inequality and more about living standards at the bottom.More soon....
(This is cross-posted from ProgressiveFix.com, the new online face of the Progressive Policy Institute, where I will be posting regularly. Give 'em a look.)To read the first part of this post, click here.Defining the CenterLet’s examine Hacker and Pierson’s definition of “the center.” When they compare activists to independents, changes in the distance from independents may be due to growing extremism among activists. However, the distance may grow without activists changing their views at all if independents change their views. So saying Republican activists drifted further away from the center than Democratic activists may misstate what occurred; independents may simply have drifted toward Democratic activists over time without activists drifting anywhere. It’s also possible that Republican activists have grown more extreme, which has pushed independents closer to Democratic activists’ (unchanged) views.
Furthermore, secular changes in ideology over time can move people from the independent category into Democratic and Republican camps and vice versa, making it difficult to say whether the changes identified indicate that activists (or independents) are changing their views, or that it’s just flows into or out of the parties that is changing. If one of the parties looks more or less extreme, it could simply be that people who would have called themselves independent in the past are now identifying with one of the parties, making the leftover independents look somewhat more extreme in the opposite direction.
Rather than compare activists to independents, why not simply measure how far they are from the midpoint of the ideology scale? When one does so, one obtains the graph below.
By this measure, which avoids all of the problems with using independents as a reference point, the change in extremism among Democratic activists looks exactly the same as the trend for Republican activists. Once again, Republican activists look more extreme in any year, and this time (not shown) this remains the case when one looks at the unsmoothed data points.
A Better Way to Measure Ideology
There is also a problem with Hacker and Pierson’s measure of ideology. If we want to know whether party activists have become ideologically more extreme over time, we should use as pure a measure of ideology as possible. The measure Hacker and Pierson use, however, conflates ideology with tolerance and empathy because it is based on questions asking how warm or cold one feels toward liberals and conservatives. It could be that Democratic activists are simply more tolerant of their opponents than Republican activists rather than being more centrist. One can feel warmly toward a group without identifying oneself with it.
A better measure of changing ideology among party activists would be to look directly at changes in self-identified ideology. The NES asks respondents to place themselves on a 7-point scale ranging from extremely liberal to extremely conservative. Here, then, is a final chart showing trends for activists in each party, with ideology measured as the distance of activists from “4” – the midpoint of the seven-point scale. The actual data points are connected and the smoothed trends are shown as black dashed lines. It should be noted that this chart is based on even smaller sample sizes than Hacker and Pierson’s, so I show the margin of error for the data points as dashed vertical lines. I also omit off-year elections to make the chart less noisy.
This chart confirms that Republican activists more often than not have been more extreme than Democratic activists, though the two groups were statistically tied in 1972, 1976, 1992, and 2004. There is a clear trend toward greater extremism among Republican activists. Among Democratic activists, there was little consistency between 1972 and 1998, but they appear to have moved to the center in 2000 and 2002 before jumping up to the level of Republican extremism in 2004.
Finally, there is the claim by Hacker and Pierson that Democratic activists are more centrist than other Democrats. In my results, this was not true in 2004 whether one used the thermometer index or the self-identified seven-point ideology measure and was not true in 2002 unless one used the seven-point measure (which Hacker and Pierson did not). Regardless, none of the differences between the two groups – in my results or theirs – are statistically significant due to the small sample sizes.
In sum, Republican activists have generally been at least as extreme as Democratic activists and often more so, though not in 2004, which makes the Republican pattern seem less worrisome. Furthermore, while in 2002 it looked like Republican extremism had increased and Democrats had become more moderate, by 2004 Democrats had completely caught up to Republicans. Republican and Democratic activists were equally far from the center in 1972 and in 2004, so the shift was of the same magnitude for both. And there’s no reliable evidence that Democratic activists are more moderate than other Democrats.
The Bush administration and the Republican Congress may have used various tactics in order to pass an agenda that lacked strong support. But they were not “off center” if that phrase is taken to mean that their agenda was outside the bounds of what the public supported. Or more specifically, where Republicans succeeded, their agenda was not out of bounds. Hacker and Pierson downplayed the extent to which Republicans had to reach out to the center in what they did or did not favor. Education spending, for instance, increased more under Bush than under Clinton, in a nod to “compassionate conservatism.” Furthermore, where Republicans truly moved off center, they failed, as with Social Security privatization. And of course, 2006 and 2008 happened.
(This is cross-posted from ProgressiveFix.com, the new online face of the Progressive Policy Institute, where I will be posting regularly. Give 'em a look.)OK, to review the debate so far: I wrote a post suggesting progressives might want to think twice before jettisoning the filibuster. Ed thought twice and said, yup, still want to get rid of it. Ezra did the same. I wrote another post saying, oh well whatever nevermind and tried to shift the subject to polarization being the real problem. I said I’d follow up about whether increasing polarization has been a one-sided affair. Crickets chirped. All hell broke loose on the health care reform front. And here we are.
So….one-sided polarization….Ever since Jacob Hacker and Paul Pierson’s Off Center, all good progressives know that the growing political polarization has been one-sided, with Republicans pulling public policy “off center” through various nefarious means. Right?
Well….yes and no. Hacker and Pierson argued that, as of 2005, Republican activists and legislators had grown more conservative, but Democratic activists and legislators had not grown more liberal (and had even moved to the right themselves in some regards). Along with this shift, Republicans had developed effective strategies to move public policy further rightward than the typical voter preferred.
Since the rightward shift of Republicans occurred during a period in which Hacker and Pierson showed the distribution of self-identified ideology had not changed, the implication was that the electorate was being deprived of the more progressive policies that it desired. But a closer look at their data and analyses shows that while the increase in polarization among legislators has occurred disproportionately among Republicans, the evidence hints that this is because it proceeded from a Nixon-era Democratic Congress that was well to the left of the electorate.
Rather than refuting the idea that policy reflects the preferences of voters in the middle (the “median voter theorem”), as Hacker and Pierson claimed, the evidence actually bolsters this view. Correcting their claims is important if progressives are to govern effectively. Republicans did not simply pull public policy to the right of where Americans preferred, and now that Democrats are back in control of Congress, progressives should not assume that the median voter is leftier than she really is.
Why Off Center Is OffTo argue their case, Hacker and Pierson turned to scores created by Keith Poole and Howard Rosenthal that put members of Congress past and present on a common scale measuring ideological position. Hacker and Pierson report that the polarization of Congress between the early 1970s and the early 2000s was almost entirely due to growing extremism among Republicans. Democratic legislators had not moved nearly as far from the center. Because of the increasing conservatism of Republicans, Congress was, in the early 2000s, far to the right of the median voter, who had not grown more conservative over time. But Hacker and Pierson’s account is flawed.
Consider the Senate.* Poole and Rosenthal’s scores, using every vote by every member of every Congress through the 108th Congress (which ran from 2003 to 2004), indicate that the “center” as of 2003-04 was typified by northeastern Republicans such as Lincoln Chafee, then-Independent Jim Jeffords, and William Cohen; Arlen Specter (now, of course, a Democrat); and by red-state Democrats such as Ben Nelson and John Breaux. In 1971-72, the median senator had a score of -0.056, equivalent to Ben Nelson’s score in 2003-04. By 2003-04, the median senator had a score of 0.061, equivalent to Arlen Specter in 2003-04.
This small change in the median of the Senate as a whole only hints at the fact that, as Hacker and Pierson claim, Republican senators did move farther ideologically than Democratic senators. The evidence that Hacker and Pierson presented describes how the median in one year compared with then-recent senators’ scores. In the early 1970s, according to Hacker and Pierson, the median Republican senator lay “significantly to the left of current GOP maverick John McCain of Arizona—around where conservative Democrat Zell Miller of Georgia stood” [where the references to McCain and Miller are to their 2003-04 scores, italics in the original]. The median Republican senator’s score then “doubled” by the early 2000s so that it sat “just shy of the ultraconservative position of Senator Rick Santorum.”
These descriptions do not quite reflect what the Poole-Rosenthal scores show. The median Republican senator’s score in 1971-72 was equidistant between McCain in 2003-04 and Miller in 2003-04, not closer to Miller, and it was just as close to McCain as the median Republican senator’s score in 2003-04 was to Santorum.
This claim also raises a technical issue. The Poole-Rosenthal scores are not ratio scales with a meaningful zero point. The distance between 0.2 and 0.4 is supposed to be the same as that between 1.2 and 1.4, but 1.2 is not “six times as conservative” as 0.2, because a score of 0 does not indicate the complete absence of conservatism. The zero point is completely arbitrary. The doubling from 0.2 to 0.4 would become an increase of just 50 percent if we added 0.2 to all of the scores (from 0.4 to 0.6). We cannot know whether Republican senators grew twice as conservative between the early 1970s and the early 2000s. Indeed, the phrase “twice as conservative” has no obvious meaning.
More to the point, Hacker and Pierson’s interpretation of these results is an even bigger problem. Rather than the Republican Party drifting ever rightward (the whole time increasingly “off center”), if the Democratic Party was “off center” in the early 1970s, then the movement among Republicans could be interpreted as a restoration of an equilibrium reflecting voter preferences. This is exactly what appears to have happened.
First of all, the medians for the 2003-04 Senate were 0.379 and -0.381 for Republicans and Democrats – essentially identical. That means that after this great rightward shift by Republicans, the parties were equally “extreme” by historical standards. Furthermore, the median Democratic senator in 1971-72 wasn’t much less extreme than the median senator from either party in 2003-04.
Second, at least in terms of self-identification, the ideological distribution of Americans was unchanged over this period, with roughly twice as many people calling themselves conservative as calling themselves liberal.**
Taking these facts together – a rightward shift by Republican legislators, an end state where Democrats and Republicans are equally “extreme”, and an ideological distribution among voters that was static over the period (and right-leaning) – the conclusion that best fits is that the Democratic Congress of 1971-72 was off center rather than the Republican Congress of 2003-04. The median Republican became more extreme over time, but that was because Congress became more representative of the electorate, not less. The story on the House side is much the same, except that the median Republican was a bit more “extreme” than the median Democrat by 2003-04 (although no more extreme than the median Democrat was in 1971-72).
Comparing the ActivistsHacker and Pierson also argue that Republican activists grew more extreme while Democratic activists became less so (becoming even less extreme than Democrats in general), but these claims are also problematic. Hacker and Pierson began by defining an activist as someone who self-identifies as a Democrat or a Republican and who participated in three out of five election-related activities asked about in the American National Election Studies. They measured ideology using a combination of two “thermometer” items – one of which asks respondents how warm or cold they feel toward liberals and one inquiring about conservatives. These scales range from 0 (cold) to 97 (hot). (The scale ends at 97 rather than 100 because in some years, the NES used codes 98 and 99 as missing value codes.) The liberal score is subtracted from 97 (so that high numbers then signify cold feelings) and then added to the conservative score. This number is divided by two, 0.5 is added to it, and the decimal is dropped. The resulting measure ranges from 0 (extremely warm toward liberals and extremely cold toward conservatives) to 97 (extremely cold toward liberals and extremely warm toward conservatives).
To determine how far activists drift from the center, they compared the activist scores on this index to the scores for independent voters. The distance from independents is expressed in percentage terms (e.g., 10 percent more conservative or liberal). Hacker and Pierson plotted the average distance from independents for Republican and Democratic activists and then “smoothed” the trends by imposing curves to describe them. The result is a graph that I replicated, more or less:
The graph shows that Republican activists were more extreme than Democratic activists to begin with, that they became more conservative over time, and that after becoming more liberal, Democratic activists tacked back toward the center. The first important thing to note about this graph is how much the nice, smooth lines depend on fitting the data points to a quadratic equation. The original data – without the smoothing – looks much messier:
The upward trend among Republican activists is still readily apparent, but the trend for Democratic activists no longer points toward moderation. The bouncing around is partly due to different turnout patterns in off-year elections, but also a result of statistical noise, as the sample sizes for each group are less than 70 – and as low as 18 – in each year. Furthermore, Republican and Democratic activists are statistically the same distance from the center for much of the period between 1968 and 1992. To illustrate further how deceptive the smoothed trend lines can be, look what happens to them when 2004 data – which was not available when Hacker and Pierson created the graph – is added:
The Republican line hardly changes, but now Democratic activists appear to grow steadily more liberal. It still appears as though Republican activists drifted from the center more than Democratic activists did, and Republican activists look more extreme in all years.
OK, take a breather. Tomorrow I’ll wrap up with some revealing evidence about how Hacker and Pierson’s definition of “the center” affects these analyses of political activists.
To read the second part of this post, click here.
* Following their recent book, Polarized America (McCarty, Poole, and Rosenthal, 2006), I use scores on Poole and Rosenthal’s first DW-NOMINATE dimension (for details, see http://polarizedamerica.com/ and http://www.voteview.com). Hacker and Pierson report using “d nominate” scores, but these are only constructed through the 99th Congress, so I am inclined to believe that they too used the first DW-NOMINATE dimension scores.
** Hacker and Pierson (2004), page 38. Hacker and Pierson cite ANES data. According to Gallup data showing self-identified ideology, the breakdown among Americans as a whole in 2004 was roughly 20 percent liberal, 40 percent moderate, and 40 percent conservative (Wave 2 of the June Poll, Question D10). In 1972, it was 25 percent, 34 percent, and 37 percent (Poll 851, Question 14).