Supporters of income tax cuts have been touting a recent poll that purports to show strong support for tax cuts in Oklahoma. A telephone survey of 500 registered voters were asked four questions about tax cuts, tax credits, and the size of government. The discerning consumer of public opinion research should be skeptical about this poll’s conclusions for two reasons. First, the survey is delicately phrased to elicit attitudes about tax cuts ‘in a vacuum’ and leaves too many influential factors out of the questions entirely. Second, the pollster’s report relies heavily on results within sub-samples (e.g. party, income) – with margins of error that reach double-digits for some groups.
The Sooner Survey poll, conducted by Republican lobbyist and consultant Pat McFerron, asks voters if they favor ‘eliminating some tax credits‘ to reduce the top tax rate, but it doesn’t tell them which tax credits. It’s likely that when asked directly about popular credits slated for elimination – including the child tax credit, child care tax credit and sales tax relief credit – many respondents would answer this same question very differently. The same applies when the poll asks respondents to agree or disagree with the statement: ‘I would favor cutting government programs and services so the savings can be passed along to taxpayers in the form of a tax cut.’ The key to an accurate answer is to present an accurate choice. How would the answers change if the poll had posed instead, ‘I would favor larger class sizes so the savings can be passed along to taxpayers in the form of a tax cut, ‘ or ‘I would favor fewer police officers..,’ etc?
While the poll reported an overall margin of error of 4.3 percent, Sooner Survey’s own reporting and subsequent press lean on results within sub-groups of the larger sample, where error margins are markedly higher. The farther you drill down into the 500 respondents, the higher the margin of error. The margin of error is important to interpreting the poll because it indicates precisely how similar this small sample is to the population in Oklahoma as a whole, some two million registered voters. For instance, the pollster writes in CapitolBeatOK:
By a better than two-to-one margin Oklahomans favor cutting the state income tax to 4.75%. Most significant groups top 50% in their support. It is supported by men (52% favor vs. 25% oppose) and women (51% favor vs. 22% oppose) urban residents (53% favor vs. 24% oppose) and rural denizens (52% favor vs. 24% oppose).
The report implies that there is landslide approval for a tax cut. Yet even when worded in the most favorable way possible, after you account for margin of error, the poll did not find conclusive majority support for cutting the state income tax rate. Standard statistical controls eliminate the slim majority in favor of tax cuts within each demographic and public opinion looks more like a dead heat than a landslide.
Additionally, half of the survey used a split sample (albeit in a well-intentioned and methodologically-sound attempt to test alternate question-wording) with just 250 respondents and a margin of error of +/- 6.2 percent. When McFerron drills down into this abbreviated sample, the poll’s error margins grow from cracks to craters. He writes:
Even among Democrats, however, a plurality favors cutting government programs and services. Again, we see the real divide being between those Democrats with an unfavorable impression of the President (60% favor vs. 28% oppose) being much more for cutting than are those with a favorable impression of Obama (34% favor vs. 55% oppose). Among anti-Obama Democrats there is actually more supportive intensity (50% strongly favor) than among Republicans (40% strongly agree).
About half of registered voters are Democrats, so Democrats would constitute roughly 125 of the respondents in the sample. Assuming for the sake of argument that half of Oklahoma Democrats are ‘anti-Obama’, that means this sample is comprised of just 62 people and has a margin of error of about +/- 12.5 percent.
According to GALLUP, it’s a truth universally acknowledged that Americans approve of cutting spending in the abstract, but disapprove of cutting specific programs. This poll fails precisely because it ignores this maxim. Unless the public is presented with the flip-side of the fiscal debate, and also asked about support for cutting specific programs to pay for these tax cuts, the results will be shallow and predictable.