“Those who ignore statistics are condemned to reinvent it. Statistics is the science of learning from experience.” – Bradley Efron
This November’s statewide ballot promises to captivate all Californians interested in direct democracy (notwithstanding any secret, diabolical, perfectly legal plans to block the 2014 initiative fest). The Secretary of State webpage shows that, as of February 24, 46 California initiatives and referenda so far have been cleared for circulation to collect signatures. Still another 8 are pending at the Attorney General’s Office.
Obviously, not all of these initiatives will survive the circulation process, and some of the initiatives filed are merely variations on the same theme. But potentially dozens could appear in front of voters. With possible subjects ranging from minimum wage to MICRA, from cigarette taxes to charter schools, and from transportation to term limits, there might literally be something for everyone.
As the November ballot takes shape over the next couple of months and thoughts turn to which initiatives will pass and which will fail, a natural question becomes: can recent California history provide any substantive insight regarding potential winners and losers?
The answer is yes. (If it weren’t, this would be an awfully short blog posting.) Specifically, by using statistical analysis.
Statistical analysis is now being used to predict everything from the profitability of individual stocks to the future performance of baseball players and teams – see Moneyball as a prime example. In the political realm, there are even regression models designed to forecast presidential outcomes (although some of these models are fraught with technical issues).
I have developed a statistical prediction model for California ballot measures based on information gleaned from elections going back to 1998. The model sifts through data on California’s electoral landscape (such as economic conditions, demographic shifts, and party registration) as well as data unique to each ballot measure (such as policy areas and campaign spending). The model is then able to estimate a measure’s probability of winning.
Not surprisingly, the nice thing about ballot measures in California is that there are a lot of them – 161 over the past 15 years. Which means more data to analyze and statistical estimates that are more precise.
While it is a little early to give predictions regarding the fate of individual initiatives, the model can still provide insight into the general tendencies of the California electorate. Here are a few interesting trends I’ve uncovered:
California voters are deeply skeptical of initiatives. Recall that under California law, one or more private individuals can work to place initiatives on the ballot, while the Legislature maintains the ability to place constitutional amendments, statutory changes, and bond measures on the ballot. My model shows that all else being equal – meaning controlling for things like the measure’s subject matter and the amount of money spent for or against – the average initiative has only a 27 percent chance of winning. By contrast, the average non-initiative has a 63 percent chance of winning.
Why the huge disparity? While it’s hard to say exactly, one possible explanation is that voters fear attempts to use the ballot box to confer special benefits to narrow interests. But there could be other explanations as well.
Of course these dire probabilities do not mean that all hope is lost for initiatives and their backers. Campaign committees in Sacramento and across the state work tirelessly to ensure that their own measures end up anything but “average.”
California voters do not tire easily. If you were thinking that having a crowded ballot might spell trouble for the slate of measures, well think again. The data does not show any statistically significant relationship between the length of a ballot and the number “No” votes or “Non” votes (i.e., where voters simply do not mark a vote) cast on ballot measures. (In fact, the data does not even show a statistically significant relationship between the length of an individual measure, as measured by the number of words, and the number “No” or “Non” votes.)
There are actually precedents for having crowded ballots in California. In March 2000, voters were greeted with 20 ballot measures, and in November 2004 it was 24.
One possible explanation for this finding is that voters are committed. That is, while turnout might vary from election to election, those Californians who actually do go through the hassle of getting to their polling place (or obtaining their absentee ballot) make the most of their time and effort.
(Now, if we hit somewhere in the neighborhood of 36 ballot measures this November, I might be inclined to back off my assertions….)
Money matters, but how much it matters depends. Proponents generally spend about $12.4 million on the average ballot measure, while opponents generally spend about $8.4 million. As a result, the average ballot measure (as distinguished from the average initiative) has roughly a 41 percent chance of success. The model shows, not surprisingly, that proponents can increase the likelihood of passage by collecting more cash and upping their spending, all else being equal. Here, doubling expenditures increases the probability of success to 55 percent.
But rarely is all else equal when it comes to campaign spending. Opponents also get a say in the matter. Additional spending by opponents can undo the work of proponents and drive down the likelihood of passage.
While statistical models can provide a lot of insight, admittedly there’s as much art as science behind what it takes to get ballot measures to pass. For example, this type of statistical analysis cannot account for the creativity – or lack of creativity – of a measure’s public relations firm or pollsters. So statistical models should be used as complements to, rather than substitutes for, other types of analysis.
That said, California has a long history of direct democracy. It’s useful to learn from this experience.
Dr. Justin L. Adams is the President and Chief Economist of Encina Advisors, LLC, a Davis-based research and analysis firm.