Starting today on Good Morning America, ABC News is reporting a new measure of public discontent in this country, something we're calling America's Frustration Index. The bottom line: Fueled by political and economic discontent alike, it's running high.
We've created this index by combining four central measures in public attitudes in ABC News/Washington Post polls – views of the president's job performance, ratings of the national economy, anti-incumbent sentiment and dissatisfaction with the way the government's working.
Each of these components correlates with broader public sentiment; they're closely aligned, for instance, with views on whether the country is headed in the right direction or seriously off on the wrong track. They also correlate with two key election outcomes – the rate at which incumbents are reelected and the loss or gain of House seats by the incumbent president's party.
We're computing the Frustration Index on a scale of 0 to 100, with higher numbers indicating greater political discontent. (It's produced by subtracting the percentage of positive answers to each of the four questions from the percent negative, adding the results and dividing by four; then, to produce the scale, adding 100 and dividing by two.)
Frustration Index: A History
Given the absence of past polls in which all four index questions were asked together, we've computed previous index numbers since 1992 using instances when all four questions were asked within at least three months of one another. We've also used other, highly correlated data to estimate the index at the time of each presidential or midterm election.
The Frustration Index started this period at 73 out of 100 in spring and fall 1992, as economically stressed voters prepared to deny then-President George H.W. Bush a second term. While it then eased a bit, the index was a still-high 63 just before the 1994 midterm elections, when – frustrated by the slow pace of economic recovery – voters handed control of the House and Senate to the Republicans for the first time in 42 years.
As the economy picked up steam, frustration eased. The index fell to 50 in 1997 and dropped to its low, 39, in 1998 – when, not coincidentally, the re-election rate of House incumbents was its highest across this period, 98.3 percent. (The re-election rate was its lowest, 88.3 percent, in 1992, when, again not coincidentally, the Frustration Index was very high.)
Frustration Index and 2010 Midterm Elections
A long rise in frustration began after the 2000 election, first with the collapse of the dotcom bubble, then growing dissatisfaction with then-President George W. Bush, the unpopular war in Iraq and ultimately the Great Recession. We estimate an index of 62 at the time of the 2006 election, when the Democrats retook control of Congress; it soared from there to its record high, an estimated 80, in fall 2008, as the economy fell into the abyss and voters removed the GOP from the White House.
The index stands at 67 percent in our latest poll, completed this past Sunday, the same place it's been all year – and its highest in available data since fall 2008, and before that, 1992.
GROUPS – The index is dramatically different across population groups – 51 among Democrats, for instance, rising to 73 among independents and 82 among Republicans. Similarly, among registered voters who currently favor Democrats for Congress, it's 53; among those who favor Republicans, it's 84. Underscoring the risk to all incumbents, the Frustration Index among swing voters – likely voters who say their party preference may change – also is quite high, 73.
The gaps extend to broader views; among people who say the country's headed in the right direction, the index is 46. Among those who say it's seriously off on the wrong track, it's 81. Among those who say the economy's in good shape, 24; poor shape, 82. And among those who approve of Barack Obama's job performance it's 47; among disapprovers, 91.
It's important to note that, in election terms, the index is "postdictive," to use Edward R. Tufte's term in his 1974 classic, "Data Analysis for Politics and Policy." It does not predict the future, in which other, unforeseen factors can come into play.
There's a particular reason for caution this year: The divergence between public views of current economic conditions, which are dreadful; and expectations of the economy's future, which have improved lately, to their best in six years, as we reported in mid-May.
If we swap out ratings of the current economy in the Frustration Index and replace them with economic expectations, the index is lower, 58 not 67, an unusually large gap between the two approaches in available data.
Both versions – current economic sentiment and economic expectations – correlate, retrospectively, with incumbent re-election and with a president's party's House gains or losses. The divergence this year makes economic expectations a significant wildcard to keep in mind as the midterm elections approach.
The Frustration Index
A note on correlations: The individual items in the Frustration Index correlate with one another at a range from .54 (presidential approval and support for incumbents) to .92 (presidential approval and satisfaction with how the federal government is working). (A correlation is a measure of data congruence, in which 1 means the data move perfectly in tandem, -1 means they move perfectly in opposition and 0 means they move independently.) The Cronbach's Alpha for the index is .88, indicating its components measure the same underlying construct. The index correlates with views that the country's on the wrong track at .89; its estimated values in midterm elections since 1994 (note the small number of cases) correlate with losses by House incumbents at .81 and with losses for the president's party at .72. Index numbers for elections were estimated with a formula obtained by regressing available index numbers on another index, created by combining presidential approval and ABC's weekly Consumer Comfort Index. The indices correlate at .99.
A full list of available index points follows. Those given with dates were assembled from existing data; those marked as election data points were estimated via regression analysis.