The Exit Polls and the Jewish Vote

Almost within hours of the release on Wednesday morning of summaries of the national exit polls, conducted with voters across the county on Election Day, I received several gloating emails from liberal Jewish acquaintances, pointing to one specific result within the exit poll data: namely how Jewish voters within the national sample, had voted in the races for the U.S. House of Representatives. That sub—sample of just over 200 people who self—identified as Jewish voters (about 2% of the total survey sample), reported that they had voted 87% for Democrats 12% for Republicans.

In the 2004 Presidential election, several surveys of Jewish voters indicated that approximately 25% of Jews had voted for President Bush.  The e—mailers this week argued that Jews had come home to their natural base within the Democratic Party, and that an ad campaign run by the Republican Jewish Coalition (RJC) to alert Jewish voters to diminished support for Israel among leading Democrats and major figures on the left (e.g Jimmy Carter, Michael Moore, Cindy Sheehan, John Conyers), had failed.

The 87% figure was quickly incorporated as gospel in articles on the election in the New York Times and many Jewish publications, as if this survey of 200—plus Jewish voters were in fact a reliable indicator of Jewish voting patterns in the recent election.  One of the  clarion calls lauding the survey results came from the National Jewish  Democratic Council (NJDC), a rival to the RJC.  Interestingly, the NJDC had condemned as unreliable a survey of Jewish voters from the 2002 election which showed increased support for Jews among Republicans, because of the small sample size: 253 in that case.

But 200 Jewish voters was plenty enough for the NJDC this time. It is not clear if hypocrisy or ignorance is the appropriate way to classify these remarkably inconsistent reactions to the two surveys by the NJDC.

The same day the national voter survey data was made available, the RJC released a survey of a much larger sample of Jews conducted by Arthur J. Finkelstein & Associates in three states and Congressional districts with close races this past week. The sample consisted of 400 Jewish voters in New Jersey (US Senate race won by Democrat Bob Menendez), 300 Jewish voters in Pennsylvania 6 (a close House race for a district including mostly western suburban areas of Philadelphia won by Republican Jim Gerlach) and 300 Jewish voters in Florida 22 (a close House race in a district covering mostly the eastern portion of Palm Beach and Broward Counties, won by Democrat challenger Ron Klein).  The two House districts both have a substantial Jewish population, especially Florida 22, and New Jersey is approximately 5% Jewish, tied with Maryland for the 2nd highest Jewish population percentage after New York State. The RJC survey suggested that the 2006 Jewish voting pattern closely resembled that in 2004:  26% support for Republicans in the House races, and 27% in the Senate contest (there were also Senate races in Pennsylvania and Florida this year).

The Finkelstein survey of 1,000 Jewish voters interviewed both older and younger voters, and members of the different Jewish branches in America — reform, conservative and orthodox.  Age and the branch of Judaism, both seemed to matter a lot in the 2004 survey results, and they did again this year: the more often a Jew attended synagogue, the more likely he or she was to vote Republican.  Similarly, younger Jews, and in particular younger Jewish males, were much more likely to vote Republican. Orthodox Jews were more than twice as likely to vote Republican as Reform Jews.

None of this is surprising. The degree of religious affiliation has been positively correlated with the likelihood of voting Republican for other religious groups in the country as well. Over time, changes in the composition of the Jewish community will impact how Jews vote. The trends are likely to favor Republicans.

Orthodox Jewish women have a much higher birth rate than less observant Jewish women.  Twenty percent of Jews under age 5 are now in Orthodox families, while the total Jewish population is 10—12% Orthodox. This means that in future years, the Orthodox percentage of the Jewish population will increase, and their population will comprise a younger  group than less observant Jews, who will decline as a percentage of the total Jewish population. Both trends — higher percentage of Orthodox, and the average or median age disparity — are likely to lead to a higher Jewish support level for Republicans in the future.

Both factors demonstrate the importance of surveying enough Jewish voters to obtain a sample of Jewish voters which would include the various sectors of the Jewish population. Such a survey will achieve more reliable results than a survey that sampled only 200 Jewish voters nationwide, where the Jews surveyed might not represent a random sample of the Jewish voting population, and might include a disproportionate number of Jews from a particular location (state or city), age group, religious affiliation, or political identification.  There is simply no basis for concluding that the small sample of just over 200 self—identified Jews in the exit poll survey was a random sample, or representative of a national  population of over 2 million Jews who probably voted this year.

Jews make up just under 2% of the American population. However, they are over 2% of the voting population. Jews are among the oldest of the various ethnic, religious or racial groups in the county, and have a much higher than average turnout rate than the national average percentage of registered voters who actually vote.  The Jewish share of the national voting percentage is often 50% higher than their share of the population due to these two factors, in other words, 3%, rather than 2%.

It would seem that a sample of 200 plus Jewish voters where Jews were 2% or less of the national sample is a serious under—representation of the likely number of Jewish voters who would show up in a truly random survey of voters nationwide (perhaps 3%).

So who was missing?

Think about this in another way: is it more likely that a Chassidic Jew from Williamsburg in Brooklyn or a very liberal  Jewish voter from the upper West Side of Manhattan would walk up to an exit pollster and agree to be interviewed? Along the same line, how many Amish voters were in the national exit poll survey?  I suspect zero.

There have been several studies suggesting that liberal voters are more willing than conservative voters to agree to be interviewed in exit polls. Some of this likely relates to suspicion of media bias by conservatives and some to a history of exit polls overstating the Democratic vote percentage in recent elections (as they did again this year, particularly in Arizona, Montana and Virginia, even if overall they were closer to the actual results than in 2004).

One does not have to have a PhD in statistics to know that this survey of  just over 200 Jewish voters within the larger national exit poll survey does not provide enough information on the survey sampling technique, or who was interviewed, to judge whether the results with regard to the Jewish voters have any reliability, and can be extrapolated to reveal a pattern in a much larger Jewish voting group of perhaps 2 to 2.5 million voters nationwide. The margin of error for such a small sample size is a lot higher than for the RJC survey, with its much larger sample size (1,000), even assuming that the exit poll data and the RJC study are both random samples of the population being surveyed.   The Finkelstein survey for the RJC had a margin of error (confidence interval) of plus or minus 3.1% with a 95% confidence level, a  standard error measure for a survey with 1,000 included.  It is a reliable indication of how Jews voted in races in the competitive states and Congressional districts surveyed.

One can also test the Jewish voting pattern by looking at the actual vote totals in particular areas of high Jewish voter concentration. In Illinois' 10th district, Republican Congressman Mark Kirk withstood a strong challenge from Democrat Dan Seals and won by about 7%. The overall district is about 20% Jewish, one of the highest percentages for any district in the country.  In Highland Park, a city in southeast Lake County, population 30,000, the Jewish percentage of the voting population is generally estimated at 65—70% of the total vote.

The precinct by precinct votes in Lake County, one of two counties partially in the district, reveals that Dan Seals, the Democrat, won the Jewish areas in Highland Park, but not with anywhere close to 87% of the votes. His victory margin was by an average of 2 to 1 or 3 to 2 in most of the Jewish majority precincts. It should be noted that in this district, the RJC ad campaign had been very visible. The Finkelstein survey indicated that Jews who had been exposed to the ads voted Republican by as much as 10% more than the average for all the Jews they surveyed. That would be consistent with the results in Illinois 10 for Jewish voters in Highland Park.

Jews remain very loyal to the Democratic Party. No one is disputing that. But the numbers this year, as in 2004, are more on the order of 3 to 1 support for Democrats over Republicans, not the 7 or 8  to 1 ratio suggested by the exit poll survey. In a very difficult year for Republican candidates, Jewish voters may have even slightly increased their support for Republican candidates as compared to 2004. Anyone who relies on the exit poll survey to draw conclusions on Jewish voting patterns this year, simply does not understand the problems with relying on this small data set (which might very well be an unrepresentative, non—random sample) or likes the results too much to care about their reliability.

Richard Baehr is the Chief Political Correspondent for the American Thinker. He has a graduate degree from the Sloan School at MIT.  He thanks M. King Deets for his assistance with this article. Dr. Deets is a former Professor of Finance at the University of Massachusetts, who earned his PhD in Finance from the University of Iowa.