Honey, It’s the Pollster Calling Again!

Back in 1988 I had the pleasure of conducting polls during the New Hampshire primaries on behalf of the Boston Globe.  The Globe had a parochial interest in that year’s Democratic primary because the sitting Massachusetts governor, Michael Dukakis, had become a leading contender for the Presidential nomination.  The Republican side pitted Vice-President George H. W. Bush against Kansas Senator Bob Dole, the upset winner of the Iowa caucuses a week before the primaries. Also in the race were well-known anti-tax crusader Jack Kemp and famous televangelist Pat Robertson.  Bush had actually placed third in Iowa behind both Dole and Robertson.

We had been polling both sides of the New Hampshire primary as early as mid-December of 1987, but after the Iowa caucuses, the pace picked up enormously. Suddenly we were joined by large national polling firms like Gallup and media organizations like the Wall Street Journal and ABC News.  As each day brought a new round of numbers from one or another pollster, we began to ask ourselves whether we were all just reinterviewing the same small group of people.

Pollsters conducting national surveys with samples of all adults or all registered voters never face this problem.  Even with the volume of national polling conducted every day, most people report never being called by a pollster.  In a population of over 240 million adults, the odds of being called to participate in a survey, even ones with a relatively large sample like 2,000 people, are miniscule.  That is still true even if we account for the precipitous decline in the”response rate,” the proportion of households that yield a completed interview.  A wide array of technological and cultural factors have driven survey response rates to historic lows over the past few years as this table from Pew shows clearly:

In 2012, fewer than ten percent of households were represented in a typical poll.  Still, even at such a low response rate, the huge size of the United States population means that any individual has only a tiny chance of being selected from a sampling universe numbers of 24 million homes.  Even for a large survey of 2,000 people, the chance of any individual household being selected is a mere 0.000008.

Those odds change drastically when we narrow the universe of eligible people to “likely” voters in an upcoming New Hampshire Republican primary.  Even including people who claim they will vote but later do not, the total universe of eligible respondents in 2012 was probably just 300,000 people.   To reach that figure I started with the total of 248,485 ballots cast in the Republican primary.  To those voters we need to add the other people who reported that they would take part in the primary but did not actually turn out on Primary Day.  For our purposes, I have used an inflation factor of 20% which brings the estimated the total number of self-reported likely Republican primary voters to 298,182 people.  I rounded that figure up to 300,000 in the tables below.

Over a dozen polling organizations conducted at least one survey in New Hampshire according to the Pollster archive for the 2012 Republican primary.  In all there are 55 separate polls in the archive representing a total of  36,839 interviews, or about 12% of the universe of likely voters.  If all 300,000 likely Republican primary voters had been willing to cooperate with pollsters in 2012, about one in every eight of them would have been interviewed.  If we choose a much more realistic response rate like ten percent, there are actually fewer cooperating likely voters than the total number of surveys collected, so some respondents must be contributing multiple interviews.  Can we estimate how many there are?

It turns out the chances a person will be interviewed, once, twice, etc., or never at all can be modelled using the “Poisson distribution.”  Usually a statistical distribution relies on two quantities, its average and its “variance,” but the Poisson distribution has the attractive feature that the mean and variance are identical.  Thus we need only know the average number of interviews per prospect to estimate how many people completed none, one, two, or more interviews.  Here are estimates of the number of interviews conducted per potential respondent at different overall cooperation rates.  At a 20 percent cooperation rate, only 60,000 of the 300,000 likely voters are willing to complete an interview.  Dividing the number of interviews, 36,839, by the estimated number of prospects gives us an average figure of 0.614 interviews per prospect.


Now we plug those values into the Poisson formula to see how many people are interviewed multiple times during the campaign.


In an ideal world where every one of the 300,000 likely primary voters is willing to be interviewed, 88.4% of them would never be interviewed, 10.9% would complete one interview, and 0.7% would be interviewed twice.  If response rates fall to  8-10%, only 20-30% of likely voters are never interviewed.

Though only a few prospects would be interviewed more than once in the ideal, fully-cooperative world, at more realistic response rates closer to what Pew reports, many people were interviewed multiple times in the run up to the 2012 primary.  If only eight percent of likely voters were willing to complete an interview, about a quarter of the prospects were interviewed twice, and one in five of them were interviewed at least three times.

We can use those estimates to see how the size and composition of the actual survey samples change as a function of response rate.


At 100% cooperation, obtaining nearly 37,000 interviews from 300,000 people means a small number, about 2,000 people, would be interviewed twice merely by random chance.  So those 37,000 interviews represented the opinions of  32,000 people who were interviewed once, and another 2,000 people interviewed twice.  As response rates fall, the total number of unique respondents, the height of each bar, declines, with a larger share of interviews necessarily coming from people interviewed multiple times.  At a 10% response rate the proportion of people interviewed multiple times just about equals the proportion of people interviewed only once.  Below that rate the proportion of people interviewed only once declines quickly.

A Sea Change in Congressional Attitudes on Surveillance

About two years ago I posted an analysis of the vote in the House of Representatives on the so-called “Amash Amendment,” an attempt to restrict the bulk collection of telephone records by the National Security Agency. Observers at the time were startled when, despite widespread opposition from the Obama Adminstration and House leadership, the amendment failed on a close vote of just 217 to 205 in July, 2013.

This past week the House had another chance to vote on the question of NSA surveillance, and this time the House voted to put new restrictions on the mass collection of telephone data. In particular, the telephone and Internet providers will retain the data and provide it only in response to government requests. This so-called USA FREEDOM Act of 2015 passed the House by a vote of 338 to 88. Perhaps the May 7th decision by the Second Circuit Court of Appeals ruling the entire NSA program illegal encouraged the lop-sided outcome.

Opponents of the surveillance programs were quick to argue that this Act does not go far enough in terms of reining in the activities of the FBI and NSA. Justin Amash himself, the author of the 2013 amendment, voted against the FREEDOM Act, arguing that:

H.R. 2048 threatens to undo much of the progress resulting from the Second Circuit’s opinion. The bill’s sponsors, and unfortunately some outside advocacy groups, wrongly claim that H.R. 2048 ends “bulk” collection. It’s true that the bill ends the phone dragnet as we currently know it—by having the phone companies themselves hold, search, and analyze certain data at the request of the government, which is worse in many ways given the broader set of data the companies hold—but H.R. 2048 actually expands the statutory basis for the large-scale collection of most data.

Even House leaders who opposed his 2013 amendment supported the FREEDOM Act.  This included both House Majority Leader Kevin McCarthy and Minority Leader Nancy Pelosi.  (The Speaker typically does not cast roll-call votes and did not vote on the FREEDOM Act.  He did choose to vote against the Amash Amendment, perhaps to indicate his strong opposition to that provision.)

I have compared the votes on the FREEDOM Act with those on the Amash amendment two years ago.


The top half of the table presents the actual votes; the lower half presents the percentages based on the Members’ votes on the Amash Amendment.  Only seven Members voted against both provisions.  Those who opposed the Amendment in 2013 voted overwhelmingly, 94 percent, in favor of the FREEDOM Act.  Most of the opposition to the Act came from the supporters of the Amendment.  Nearly two out of five of those Members opposed the Act as did Amash himself.  New Members of the House also voted heavily in favor of the Act; only eleven of the 67 new Members in the 114th Congress voted no, or just 16 percent.

Is There a “Two-Term Itch?”

Over the weekend MSNBC commentator Steve Kornacki coined the phrase “two-term itch” to describe how Americans since World War II have usually opted to vote out the party of the presidential incumbent after two terms.  He concluded that this “history” might prove as great an obstacle to Hillary Clinton’s bid for the White House as any of her erstwhile Republican opponents.  Here is his evidence:


The record is quite impressive.  Only once since 1944, when George H. W. Bush beat Michael Dukakis in 1988, has the party of a two-term incumbent retained the Presidency in the following election.

On the other hand, we are only talking about a sample of seven elections, and hardly a random sample at that.  Though Mr. Kornacki’s hypothesis seems a reasonable one we might want to test it against the full range of American history.

If there is a “two-term itch,” we should expect to see it throughout the history of Presidential Administrations.  There is no reason to think voters should tire more quickly of Presidents and their parties today than they did in 1816. In that year Republican James Monroe succeeded two-term Republican President James Madison who himself took office after the two terms of Republican Thomas Jefferson.  Voters famously returned Franklin Delano Roosevelt to office four times. So while Mr. Kornacki’s hypothesis has some merit on its face, it is easy to think of counter-examples.  Just how often have American voters succumbed to the two-term itch and how often have they stuck with the horse they know?

Before we start we have to define the precise event we are trying to count.  For instance, the election of 1940 clearly qualifies as a test of the hypothesis, as it came eight years after FDR’s initial victory in 1932.  But what about the election of 1944?  Roosevelt was in the White House for the eight years prior to that election as well.  For analytical purposes I will count as a test of the “two-term itch” hypothesis any two consecutive terms where the same political party held the Presidency.  Using that definition gives us this list of twenty-four cases dating back to Washington’s departure from office in 1790.


Clearly voters since World War II have felt the “itch” much more often than voters in elections from our earlier history.  All told there are thirteen instances where the President’s party lost the office after two consecutive terms in the White House, compared to eleven instances where the President’s party held the Executive Branch, hardly different from fifty-fifty.  That includes counting Vice-President John Adams’s succession to Washington in favor of the itch theory. Woodrow Wilson’s unlikely election in 1912 alone generates two supporting cases. Roosevelt pursued his Bull Moose dream and split the natural Republican majority allowing Wilson to take the White House in 1912.  Eight years later the Republicans were unified and easily defeated Wilson’s successor James Cox.

Of those thirteen instances that meet the “itch” theory, seven of them come from elections after World War II.  Perhaps it is something about postwar American politics and presidential elections that has led to such frequent alternation in parties.

Under what conditions should we expect to find the parties alternating in office more frequently?  The obvious answer is during periods when the balance of power between the parties in the electorate is reasonably close.  The bigger the margin between the parties, the more often the dominant party will hold the Presidency.  Look back at the early years of America when Jefferson’s Republicans held the White House for the first forty years of the nineteenth century.  The contemporary Republican Party held sway over American politics for most of the years between the Civil War and the New Deal.  Since the War, though, the parties have been much more competitive.

We have popular vote totals for all the Presidential elections in the list above starting with John Quincy Adams’s victory in 1824.  If we calculate the average margin of victory in Presidential elections before and after Truman’s victory we get:

Average Margin of Victory in Presidential Elections Listed Above

1824-1944: 13.3%
1948-2012: 6.7%

The margin of victory in postwar elections is just half that in the earlier history of the Republic (and statistically different at any conventional level).  The more frequent alternation of parties we observe since FDR reflects the closer balance of support between Democrats and Republicans.  While this does not support Kornacki’s notion of a “two-term itch” being an historical feature of American politics, it is the “history” facing Hillary Clinton as she heads into 2016.


Everyone Has a Horse in the 2016 Race

Only fifteen percent of Republicans recently polled still undecided for 2016.

Already the Presidential primary season is upon us with another crowded field of contenders for the Republican nomination.  Though we are a year away from the actual primary season, pundits have already weighed in on the prospects for Jeb Bush, Scott Walker, Ted Cruz, Rand Paul, Chris Christie, and the rest.  I was curious to see how this year’s contest compared with the race for the 2012 nomination.

Using the archive of polls at RealClearPolitics for both 2012 and 2016, I charted trends for four groups of voters.  The dark blue line represents the “establishment” candidate, Romney in 2012 and Bush in 2016.  The red line plots support for two “social conservatives,” Rick Santorum in 2012, and Santorum plus Mike Huckabee in 2016.  Preferences for all the other candidates are summarized by the light blue line, while the grey line portrays undecideds.


At this time four years ago, about two-thirds of Republicans had no preferred candidate.  Once the debates began Republicans started choosing their champions.  Romney’s support improved during 2011, but most undecided Republicans were looking elsewhere.  At various times in 2011 Michelle Bachmann, Rick Perry and Herman Cain each zoomed to the top of the polls only to fall precipitously a few weeks later.  Newt Gingrich and Ron Paul showed more staying power with campaigns that persisted until the end of the calendar. Rick Santorum’s surprise tie with Romney in Iowa, followed by victories in the Minnesota and Colorado caucuses on February 7th, made the social conservative Romney’s last rival.  His victories in the March contests sealed Santorum’s fate.


If we look at how the early polling has gone for 2016, we see one striking change from four years ago.  Most Republicans today express a preference for one of the candidates, and the already diminished ranks of the undecided have fallen sharply since the end of last year. While only a third of Republicans had chosen a candidate at this point in the 2012 campaign, today more than five out of six already have someone they prefer.

Jeb Bush has been the beneficiary of some of those decisions.  He was polling a bit behind where Mitt Romney stood four years ago but has since improved to the high teens, about equal with Romney’s support in late March of 2011.  However, like Romney found after the first debates in 2011, most Republicans now choosing candidates are gravitating toward someone other than Bush.

The religious candidates start out from a much stronger base in 2014-2015 than they did in 2010-2011.  It was only after the 2012 Iowa Caucuses that Santorum’s support rose much above three or four percent.  This time around some ten to twelve percent of the Republicans polled pick Huckabee (mostly) or Santorum.  If we add Ben Carson to these social conservatives their collective level of support rises to about twenty percent of Republicans.

The race in 2010 shows how the array of “others” can quickly collapse into a two- or three-person race after the early contests.  Assuming Bush has the resources to try and follow Romney’s path, we should expect the race to narrow to Bush, a social conservative, and the survivor of the “anybody but Bush” contest.  With so many candidates running essentially even in the early polling, we might expect to see a succession of non-Bush alternatives climb to the mountaintop in the polls before being pushed down the slope by the next contender.

Still with the candidates much better identified in the voters’ minds so far before the debates in August, this “candidate-of-the-week” pattern may not recur in the 2016 race.  The share of undecided voters fell quickly four years ago because Republicans were largely indifferent among the alternatives to Romney.  That led to their switching their support to whomever appeared to be the leading alternative to Romney at the time.   It may prove more difficult to convert a Cruz, Walker or Carson supporter into a voter for Rubio.


Comparing Precipitation in States

The parched earth I see in televised news reports from California seems a universe removed from the lingering snow drifts outside my window near Boston. Here in the Northeast it feels like we have had more than our average share of rain and snowfall over the past decade.  That perception illustrates a larger reality.  Massachusetts has seen about five inches more precipitation on average since 1970 compared to levels dating back to the turn of the twentieth century.


No such good news seems to be in the cards for California.  Its average precipitation level runs about half that of Massachusetts, and unlike the Bay State, statistical tests show no trends or significant differences between older and more recent data.  The current drought appears driven by a succession of abnormally low years starting around 2007 and some historic temperature highs over the past three years. California reached its historical low for rainfall in 2013, a miniscule 7.93 inches.  In 2014, the temperature for California averaged four degrees (Fahrenheit) above the state’s 1901-2000 baseline. Temperatures in California have been running above the historical baseline for many years now.

Inches of precipitation give us the vertical dimension of a state’s total volume of rainfall; the other needed dimensions are measured by its area.  We can estimate the total volume of precipitation that fell on a state each year by multiplying together its inches of precipitation and its total area. Americans use the acre-foot as the standard unit of measure for large volumes of water.  An acre-foot represents the amount of water required to cover an acre-sized flat surface to a depth of one foot.  That works out to 43,560 cubic feet or about 326,000 gallons.

To estimate the total volume of precipitation for an average year in California, I’ll start with NOAA’s “baseline,” California’s average for the years 1901-2000 of 22.9 inches.  I divide that by twelve to convert to feet, then multiply the result by the area of California in acres to get

22.9 inches / (12 inches/foot) (158,648 sq mi)  (640 acres/sq mi) = 2.27 billion acre-feet

of precipitation in an average year.  For much smaller but wetter Massachusetts the comparable figure works out to about a tenth that for California

44.6 inches / (12 inches/foot) (8,262 sq mi)  (640 acres/sq mi) = 236 million acre-feet

Massachusetts is one-twentieth the size of California but sees about twice as much precipitation.  These factors offset to make the total volume for Massachusetts about a tenth that for the Golden State.

This calculation gives us a rough estimate of the “supply” of water available from precipitation within a state. Normalizing this total supply by the size of each state’s population gives us a metric we can use to compare water resources across states, their acre-feet of annual precipitation per capita. If we now divide that volume by the total number of people living in a state, we can compare these values across states and time like this:


At the beginning of the twentieth century Florida had the most abundant precipitation volumes per capita and Massachusetts the smallest.  Massachusetts also had a larger population (2.8 million) than any of the other states in the chart except Texas (3.0 million).  The graph portrays how the rapid growth of the Sunbelt destinations like California, Florida, and Texas reduced the per-capita supply of precipitation.  The other three states show slower declines in the availability of water because of their lower rates of population growth.  Nebraska’s combination of a large surface area and small population gives it ample supplies of water to fulfill its citizens’ needs and those of agriculture.  Massachusetts shows that even a small state with consequently limited water resources can manage if population grows slowly enough.  The state that clearly stands out the most is California.

Per-capita precipitation in the Golden State has now declined to Massachusetts levels, but California’s entire economy and society is geared to high-demand activities from almond groves to swimming pools.  Forty percent of California’s water is devoted to agriculture.  In Massachusetts that figure is less than ten percent, and most of that goes to irrigate cranberry bogs.  On its current path, California faces some stunningly difficult policy choices in the years ahead to adapt its economy to precipitation levels like those seen in Massachusetts.

Millennials Still Eschew Guns

The 2014 General Social Survey is now publicly available, so I have updated the charts that appear in my earlier postings on gun ownership by age and “generation.”

Younger millennials still show the lowest rates of gun ownership of any group in the survey.  Older millennials look more like their GenX age peers.  As I discussed in my earlier pieces on this subject, this could be a “life-cycle” effect where people buy guns as they age, though older generations did not show such trends.

It is still the case that millennials under thirty who were interviewed in 2012 and 2014 show higher ownership rates than those interviewed in 2008 or 2010.


However the growth in ownership we saw among the youngest groups between 2008 and 2012 seems to have reached a plateau in 2014.

The Rhythm of Senate Elections

From reading media reports of the 2014 election results you might believe the nation has experienced a political change of cataclysmic proportions. Republicans won 23 of the 36 states where a senatorial election was held, enough to give them control of the Senate for the next two years. Yet we need look back only to another strong Republican year, 2010, to see nearly identical results. In that year the Republicans took 24 of the 37 states where elections were held.

Historically the parties’ shares of Senate elections have swung back and forth quite substantially with the last decade appearing unusually unstable.  Here are the results for Senate elections back to 1936:



The Democrats reached their peak in the 1964 Johnson landslide, though this election merely confirmed the Democrats dominance of the Senate “class” elected six years during the 1958 Eisenhower recession.  Republicans have won about two-thirds of the seats in half-a-dozen elections over the same period of time.  The 2014 result is quite similar to the Republican margins in 2010, 2002, 1980, 1952, and 1946.

If you look carefully at this graph, you’ll see a certain rhythm in these results, one created by the six-year length of a Senatorial term and the power of incumbency.  In fact, if we slide the graph forward six years and superimpose the results, we get this:


Now we see how the partisan split in a Senate “class” helps explain the variation from election to election.  Because incumbents have an advantage when it comes time for re-election, the partisan composition of a Senate class tends to repeat at six-year intervals.  The Republican edge in 1946 was replicated six years later when Eisenhower won the White House. Six years after that a major political shift occurs.  The largely Republican class of 1946 and 1952 was replaced with a largely Democratic class during the Democrats sweep of the 1958 off-year elections. The Republicans’ victories in 1980 constituted a similar shift for their party.

Of course the partisanship of the class facing re-election just sets the stage on which each year’s set of electoral forces plays out.  We would expect that factors like broader trends in partisanship and the state of the economy might also influence the outcome of Senate elections.  I turn to those influences in the next article.



Millennials and Guns Updated

Just about a year ago I wrote here about the decline in gun ownership among those Americans who have come of age since the Kennedy Administration. Starting with the “Boomers,” rates of gun ownership by household have fallen by about ten percent each generation. Among those coming of age since the turn of the 21st Century, the “Millennials,” only 19% reported living in a household with a gun present.

Those results were based on the responses to the General Society Survey, which has been conducted periodically by the National Opinion Research Center since 1972. My earlier results were based on the combined surveys from 1972 through 2010. Now that the 2012 installment is available, I have reproduced the results of my earlier analysis with those respondents included. The entries report the percentage of households with a gun present for each combination of generational cohort and age when interviewed.


The new data gives us the ability to estimate the ownership rate for two groups not present in the 2010 sample, but the size of each sample is fairly small. Boomers now reaching retirement age show a higher rate of gun ownership than younger people in the same cohort, but that 50% for the oldest Boomers is based on just 34 people and is not statistically reliable. A chi-squared test of ownership against age group among Boomers does not reach significance.

More disturbing perhaps is the reported 28% rate of ownership among the 53 Millennials just now reaching their thirties. That figure is eight points higher than the rate for younger Millennials, but again the difference between the age groups does not meet the usual criteria for statistical significance (p < 0.20). Still if we see continue to see increasing ownership rates among Millennials as they age, my earlier rejection of a role for life-cycle factors in the decision to own a gun may have been premature. Since all the over-29 Millennials were interviewed in 2012, their higher rate of ownership could be the result of historical events that took place between 2010 and 2012. Evidence from sources like Gallup suggest that household ownership rates spiked in 2011 after years of stagnation. FBI records of the number of background checks performed show a similar spike in purchases that year. Breaking out the rates of ownership by year of interview in the GSS shows some corroboration for those trends.
age_trends_2008-2012The reported rate of ownership in 2012 for the country as whole hovered in the mid-thirties during the Obama years. Ownership estimates for the oldest cohorts declined sharply in 2010 then rebounded in 2012. For the “New Deal” cohort, aged 83 or older in 2010, these fluctuations can be attributed to small sample sizes. That is a less plausible explanation for the “Silents” (aged 65-82 in 2010); their samples each year ranged from 196 in 2008 to 144 in 2012. I have no explanation for the variation among these older Americans.

More striking is the consistent increase in household gun ownership rates among the Millennial generation, rising from 17% in 2008 to 25% four years later. With a sample of 243 people in 2008 and 310 people in 2012, this eight percent increase in gun exposure is “statistically significant” at the 0.02 level. Future posts will look more deeply into the reasons for this rise in gun exposure.

The Virginia Gubernatorial Election

On Tuesday, the Democrats won the governorship in the Commonwealth of Virginia. Though Terry McAuliffe held a substantial lead in the polls over Republican Attorney General Ken Cuccinelli throughout most of the campaign, McAuliffe’s actual margin of victory was just 2.5%, winning 48.0% to 45.5%. The narrowing of the lead during the final days of the campaign has led to much speculation over the “meaning” of this election.

This was a race between two unpopular major party candidates to succeed a Republican governor, Bob McDonnell, whose own tenure has been rocked by influence-peddling scandals. (Both McDonnell and his wife have been indicted by Federal prosecutors since I wrote this originally.) McAuliffe, a former Chair of the Democratic Party, raised prodigious amounts of money during the Clinton Administration and was accused of shady business practices. Even left-leaning commentators like Stephanie Mencimer of Mother Jones had little good to say about the Democratic candidate for governor of Virginia:

McAuliffe represents an unseemly slice of Washington. His primary role in politics for the past two decades or more has been raising money—most notably, for the Clintons. He cooked up the idea of essentially renting out the Lincoln bedroom during the Clinton administration as a fundraising vehicle, and he smashed all previous presidential fundraising records in the process. When McAuliffe was the Dems’ top fundraiser, a campaign finance scandal besieged the Clinton White House. Coincidence? No. McAuliffe was all about pushing the envelope when it came to the political money chase.

That alone might not be enough to render him a distasteful political candidate. What’s different about McAuliffe is his brazen mixing of his campaign fundraising activity and attempts to enrich himself personally. Many of McAuliffe’s business deals have come about due to his place in the political cosmos, not because he possesses a wealth of business skill. That tangled history has linked him to a long list of unsavory characters.

McAuliffe’s Republican opponent brought his own substantial baggage to the campaign. Long a crusader for right-wing causes, Cuccinelli built a reputation by aggressively opposing abortion rights, same-sex marriage, immigration reform, and gun control. In its post-mortem on the election the Washington Post put the blame for Cuccinelli’s loss directly at the feet of the candidate himself:

Fundamentally, what caused Tuesday’s Republican wipeout was Mr. Cuccinelli himself and the record he compiled — a clear, consistent right-wing agenda forged over a decade in Richmond.

The Cuccinelli record had nothing to do with job-creation or the state’s economic well-being or alleviating deepening transportation problems, all of which are central to Virginians’ well-being. It was mainly about bashing homosexuals, harassing illegal immigrants, crusading against abortion, denying climate change, flirting with birthers and opposing gun control. A hero to the tea party and a culture warrior of the first rank, Mr. Cuccinelli lost because he was among the most polarizing and provocative figures in Richmond for a decade. That made him the wrong candidate for Virginia.

Recent polls show the levels of voter dissatisfaction with the choice between McAuliffe and Cuccinelli. In the final pre-election poll by Public Policy Polling, both major candidates were viewed unfavorably by 52% of likely voters. However Cuccinelli was by far the loser among voters who viewed both candidates unfavorably; in that 15% of the Virginia electorate, McAuliffe outpolled Cuccinelli by 61% to 16%.

What factors influenced this result?

Pundits have proposed a variety of explanations for McAuliffe’s slim victory. Some pointed to the sustained turnout of African-American voters, who constituted 20% of the voters on November 5th, a figure equal to their share of the electorate in the 2012 Presidential election. For an election being held in an “off-off” year, that is a substantial tribute to the mobilization of black voters by the Democrats in Virginia, and particularly the efforts of the Obama campaigns, that began in 2008.

Another common argument concerned the government shutdown that took place in the weeks leading up to the November election. Virginia has a large number of government workers, not only in the suburbs around Washington, but in places like Norfolk where defense industries constitute an important part of the local economy. Observers suggested that disgruntled government workers turned out for McAuliffe to express their unhappiness with the government shutdown which they, and most other Americans, largely blamed on the Republicans, particularly the Party’s Tea Party wing. The dissatisfaction with the Tea Party spilled over onto Cuccinelli with his own strong ties to that faction.

Despite these factors that advantaged McAuliffe, the margin between the two major party candidates narrowed as the election drew near. The Democrat held a single-digit lead during most of September and early October, but by the middle of that month his margin ballooned into a lead of 10-17%. The highest figure recorded for McAuliffe rather surprisingly came from Rasmussen Reports, the Republican polling firm which consistently over-estimated support for Mitt Romney during the 2012 campaign.
During the last weeks of the campaign Cuccinelli took a cue from Tea Party politicians like Ted Cruz and began to stress his opposition to the Affordable Care Act, better known as “Obamacare.” Cuccinelli had led the charge against the law in 2010 when he filed suit to declare it unconstitutional, an action that was quickly followed by similar suits from Republican attorneys general across the country. Conservative commentators have since argued that Cuccinelli waited too late to bring the ACA to the forefront of his campaign, and had he done so sooner, he would have defeated McAuliffe.

The Sarvis Vote

Another complication for understanding this election was the candidacy of Libertarian Robert Sarvis. Sarvis had once run as a Republican but left the party because he thought it was too socially conservative. He holds a rather mixed bag of issue positions, some considered left-wing like his support for the legalization of marijuana, and some from the right like his opposition to income taxes and gun control. Conservative commentators have suggested that Sarvis was instrumental to McAuliffe’s victory by siphoning votes away from Cuccinelli. Some pointed to campaign contributions to a Libertarian PAC from a highly-placed Obama fundraiser in Texas as evidence for Sarvis’s role as a stalking horse for the Democrats.

Whether Sarvis was closer ideologically to Cuccinelli or McAuliffe probably did not matter much in voters’ decision-making calculus. One constant in the Virginia polling data was how little the public knew about Sarvis or his positions on the issues. In a late-October pre-election poll by the Washington Post, fully 43% of likely voters reported knowing “nothing at all” about Sarvis, and another 35% chose the “just a little” option. Only 3% reported knowing “a lot” about Sarvis, with another 18% saying they knew “a fair amount.” That 21% of voters who claimed to know something about Sarvis stands in stark contrast to the two major-party candidates, where the comparable figures were 81% for McAuliffe and 86% for Cuccinelli. While some of the nearly 7% of Virginians who cast ballots for Sarvis on November 5th probably supported him on the issues, it seems likely that most of his voters were largely ignorant of his positions.

Evidence from the Election Results

I have subjected some of these arguments to empirical testing by examining the vote by jurisdiction as reported by the Virginia State Board of Elections for the 133 counties and cities in Virginia. The model below estimates the vote for Cuccinelli as a function of the vote for Republican Bob McDonnell in the 2009 gubernatorial election and a few demographic factors like the proportion of African-Americans and Hispanics in each jurisdiction and the fraction of the population in each locale that were employed by the Federal government. Because the government shutdown affected civilian and defense workers differently, I have estimated separate effects for both groups. I also include the vote for Sarvis to see whether his candidacy influenced the margin between the two major-party candidates.

I start with a model that includes only the 2009 vote and demographics.

Data: 133 Virginia Counties and Cities
Dependent variable: Vote for Cuccinelli

               coefficient   std. error   t-ratio   p-value 
  Constant     −0.996052     0.0633928    −15.71    9.50e-32 ***
  McDonnell-09  0.827063     0.0334505     24.73    8.75e-51 ***
  Afr-Amer     −0.0889980    0.0128968     −6.901   2.09e-10 ***
  Hispanic     −0.139432     0.0192821     −7.231   3.77e-11 ***

Mean dependent var   0.040395   S.D. dependent var   0.583154
Sum squared resid    4.267090   S.E. of regression   0.181874
R-squared            0.904941   Adjusted R-squared   0.902731

All variables are measured as "logits."

Unsurprisingly the 2009 gubernatorial vote has by far the greatest predictive power, but both the demographic factors show significant effects as well. Of particular note is that the size of the Hispanic population in a jurisdiction has a slightly larger effect than the size of the African-American population. A statistical test for the equality of the two effects shows that the Hispanic coefficient is significantly larger than the one for African-Americans.

If we now add in the measure of Federal employment in each jurisdiction we find it adds no additional explanatory power to our simple demographic model.

                coefficient   std. error   t-ratio   p-value 
  Constant      −1.04497      0.0760838    −13.73    6.07e-27 ***
  McDonnell-09   0.816998     0.0345162     23.67    1.37e-48 ***
  Afr-Amer      −0.0918772    0.0131170     −7.004   1.26e-10 ***
  Hispanic      −0.129533     0.0210656     −6.149   9.21e-09 ***
  Fed_Emp       −0.0199454    0.0172078     −1.159   0.2486  

Mean dependent var   0.040395   S.D. dependent var   0.583154
Sum squared resid    4.222768   S.E. of regression   0.181633
R-squared            0.905929   Adjusted R-squared   0.902989

While the coefficient for Federal employment has the expected negative sign, it is only slightly larger than its standard error and falls well short of the conventional p

This is a rather surprising result given the journalistic coverage of the effects of the government shutdown on the race. It turns out that jurisdictions with larger proportions of Federal employees are also ones with larger Hispanic populations (r=0.42). If we exclude the Hispanic measure from the model, the coefficient for Federal employment then becomes significant.

                coefficient   std. error   t-ratio   p-value 
  Constant      −0.829620     0.0765776    −10.83    7.57e-20 ***
  McDonnell-09   0.824725     0.0391062     21.09    1.24e-43 ***
  Afr-Amer      −0.112898     0.0143573     −7.863   1.30e-12 ***
  Fed Emp       −0.0628453    0.0178336     −3.524   0.0006   ***

Mean dependent var   0.040395   S.D. dependent var   0.583154
Sum squared resid    5.470150   S.E. of regression   0.205923
R-squared            0.878141   Adjusted R-squared   0.875307

However this model does a poorer job of explaining variations in the Cuccinelli vote than the model that includes the Hispanic measure. Additional tests separating out civilian and defense employment show no improvement in explanatory power. Observers who saw a correlation between Federal employment levels and the gubernatorial vote were likely reacting to the more significant correlation with the size of the Hispanic population.

Finally I include the vote for Robert Sarvis, which definitely has a significant and negative effect on the vote for Cuccinelli.

               coefficient   std. error   t-ratio   p-value 
  Constant     −1.21892      0.123557     −9.865    2.03e-17 ***
  McDonnell-09  0.831222     0.0330809    25.13     2.51e-51 ***
  Afr-Amer     −0.0882480    0.0127363    −6.929    1.85e-10 ***
  Hispanic     −0.138816     0.0190369    −7.292    2.82e-11 ***
  Sarvis       −0.0833661    0.0398514    −2.092    0.0384   **

Mean dependent var   0.040395   S.D. dependent var   0.583154
Sum squared resid    4.126027   S.E. of regression   0.179540
R-squared            0.908084   Adjusted R-squared   0.905212

As I mentioned earlier, most voters had little idea who Sarvis was or what he stood for. Rather than representing support for either the candidate or his policy positions, the Sarvis vote more likely represents dissatisfaction with one or both of the major party candidates.

As it turns out, I found a similar pattern in voting for the British Liberal Party during the 1960s and early 1970s. In that article, and in my dissertation research, I showed that voters who switched their support from the Conservative or Labour parties to the Liberals often had specific issue disagreements with their party of origin. Since most voters viewed the Liberals as ideologically “between” the major parties, unhappy supporters of those parties could cast a protest ballot without having to cross all the way over to the opposition. Moreover, in many cases these switchers held positions that differed from the Liberals’ own positions on the issues. For instance, voters who opposed Britain’s entry into the Common Market were more likely to move to the Liberals despite the fact that the Liberal Party was itself the most pro-Market of the three main British parties.

The theoretical basis for this argument follows from the work of American political scientist V. O. Key who argued that elections were primarily retrospective evaluations of past performance and not mandates for future policy. Seen in this light, it makes eminent sense that voters for Sarvis were more likely to come from the ranks of potential Republican voters than Democratic ones. Voters who might otherwise have supported Cuccinelli, but were unhappy about the scandals plaguing the McDonnell administration, found it convenient to vote for Sarvis rather than having to cast a vote for McAuliffe. The same pressures would not have applied as strongly to Democrats despite their mixed views about Terry McAuliffe. While some Democrats might have been turned off by his rather smarmy reputation, the pressure to “rally round the flag” and oust the Republicans probably kept many Democrats from defecting to Sarvis. As a result, the Sarvis coalition contained more likely Republicans than likely Democrats, resulting in the negative coefficient we see in the results above.

Most Vulnerable Republicans Voting against the Shutdown Resolution


This table presents those Republicans who voted against H.R. 2775, the resolution to end the government shutdown and extend the debt ceiling until February 7, 2014.  The “incumbent” column indicates whether the Member already held the seat during the 2012 election.  Eight of the twenty are newcomers, though all but one come from districts carried by Mitt Romney in last year’s Presidential election.

According to Roll Call, Michelle Bachmann has already announced her intention to retire. The remainder have made no announcements about whether they will run for re-election in 2014.