Exemptions and employment

A new study from the University of Minnesota School of Public Health about the impact of smoking bans on restaurant and bar employment is making news this week. The study examined 10 Minnesota cities from 2003-2006, tracking employment numbers and searching for effects correlated with different smoking policies (comprehensive bans, bans with exemptions for bars, and no bans). No statistically significant impact was found, so the researchers conclude that governments should pass comprehensive smoking bans without concern for economic effects. Predictably, the conclusion has been picked up uncritically by the press, and the authors hope it will influence debates over statewide bans.

The study is gated online so I wrote to lead author Liz Klein, now at Ohio State University, and she was kind enough to send me the full paper so that I could look at the methodology. I’ll get to that in a moment, but first I’d like to reiterate that employment effects are among the least important arguments against smoking bans. The real question is whether consenting adults should be free to assemble in a privately owned business to enjoy a legal product. Reasonable people can disagree about various measures to reduce indoor smoking, but to eliminate all businesses where adults can enjoy tobacco and alcohol is a blatant violation of their rights to property, assembly, and contract. Empirical questions about employment are an interesting part of the debate over smoking bans, but while discussing them we shouldn’t lose sight of how anti-liberal comprehensive bans really are.

That said, let’s get to the study.

Fortunately, the methodology is straightforward. Most research in this area focuses on tracking changes in employment in one jurisdiction before and after a smoking ban. This one takes a different approach, tracking employment in 10 cities classified by their smoking policies. Employment was measured by restaurant’s and bar’s self-reported numbers of employees, which they are required by law to send to the state. Two of the cities had no bans on smoking in bars and restaurants and served to identify any secular changes in employment trends. After adjusting the numbers into a per capita measure to account for population differences, no significant effects were found.

All of this is fine as far is it goes, but there are some important limitations. The first is that the cities in the sample weren’t randomly treated with smoking policies. Comprehensive smoking bans get passed in the cities that have the political will to pass them and thus presumably have many residents who like going out to smokefree businesses. Cities with looser bans or no bans at all likely have more relaxed attitudes toward smoking. The employment effects of local bans are therefore potentially very different from those of bans imposed statewide, in which, for example, the preferences of urban voters might trump those of rural bar patrons who’d very much enjoy a cigarette and a beer.

This isn’t idle speculation. A 2006 study (not gated) by economists Robert Fleck and Andrew Hanssen examined California’s statewide smoking ban that went into effect long after many California communities had passed their own local restrictions. Here is a summary by Michael Pakko at the St. Louis Fed:

[Fleck and Hanssen] analyzed quarterly restaurant sales data for 267 California cities over 25 years. They find that the measured impact of smoking bans differs between local bans and the statewide ban. In what the authors call their “naïve” specification that treats all smoke-free laws the same, they find a statistically significant 4 percent decline in revenues associated with smoking bans.

When they estimate the effects of the statewide ban and local bans independently, they find that the measured decline in restaurant sales is attributable to the statewide ban on cities without local bans. The measured effect of the statewide ban is nearly 4 percent, and it is statistically significant. The independent effect of local smoking ordinances is estimated to be very small and is not significant. These findings are consistent with the interpretation that locally originated smoking bans have little effect, but smoking bans that are imposed on a community by a higher jurisdiction can have a detrimental economic impact.

Fleck and Hanssen go on to uncover an important specification problem: They find that cities that adopted smoke-free laws were systematically different from those that did not. The authors find that sales growth tends to be a predictor of smoking bans, rather than the other way around. This “reverse causality” calls into question many earlier findings, and it poses problems for using data from California in drawing inferences about the economic impact of smoking bans elsewhere.

I asked Professor Klein if she thought this limitation should be considered when applying her study to Ohio, where bar owners are currently reporting declining revenues and seeking an exemption from the state. She responded:

While randomization is an ideal option to deal with the potential for confounding, a quasi-experimental design, as we have implemented, is the next best option. We selected comparison cities (by size and type of community) in order to best account for time trends in effect at the time of study (2003-2006).

In other words, the limitation is real, and the applicability of the study depends on one’s confidence that the two small control cities are an adequate substitute for randomization. Given the inherently non-random distribution of local smoking bans, I suggest a caveat ought to be included in discussions of statewide exemptions.

The second major limitation is even more serious, and that is that the study only tracks effects at the community level. Individual businesses could be hit drastically hard by a ban and not show up in the data if local restaurants and bars in general are performing well (and remember, a successful hospitality industry may be one predictor that a community will pass a ban). Klein and the other authors acknowledge this in their paper:

An important limitation in the use of aggregated data is that we are able to estimate an overall average, but we are not able to determine differential effects at the business- or neighborhood-level.

I mentioned this to Klein as well. She replied:

Our study does not make claims about individual businesses, as we had data at the community-level only. However, policies are enacted at the community level, so it seems that evaluation of policy effects at the community level are appropriate. Ultimately, decision-makers have to weigh the scientific and economic evidence to make the best decisions for the health of a community.

Voters and legislators consider exemptions to smoking bans precisely because they realize that policies that are good for the community at large might be disastrous and/or unfair to certain businesses within it. To say that community-level effects are the only thing that matter is to dodge the question. Essentially, it is to tell people put out of business by a ban to suck it up because on the bright side their competitors are doing well. Fortunately most people are more tolerant and compassionate than that and so exemptions to smoking bans may continue to pass.

Two other limitations of the study also come to mind. The first is that job numbers are an imperfect measure of the economic effect on employees. They tell us how many people are working, but they don’t tell us how their hours on the job or tip revenue have been affected. Employees in some bars included in the study might have been made worse off by anti-smoking policies and we wouldn’t know it from the data.

The second is that 2003-2006 were good years for the hospitality industry. The economic picture is very different in 2008-2009. Bars in the study period were likely able to absorb the shock of smoking bans more easily than bars today.

In summary, this new study is interesting and credible. It (and others) suggests that ban opponents sometimes overstate the economic case against smoking bans. Its application is limited, however, and says nothing about the potentially disastrous impacts of smoking bans on individual businesses and employees. Policy makers should continue carving out exemptions to smoking bans for the sake of business owners, their patrons, and the people who choose to work for them. A free society can and should tolerate a diversity of smoking options.


Just give it to the Lizard People

Writing in The New York Times, mathematics writer Charles Seife says the Minnesota Senate race is too close to call:

So in some precincts, we have not just a recount but a re-recount. Both auditors and recounters were hypervigilant to possible sources of error, and yet they disagree on their tallies by about 20 thousandths of a percent.

In an ordinary race, errors this tiny wouldn’t be a problem. But the Coleman-Franken race is so close that this error rate is more than double the margin between the two camps. And that’s just taking into account the precincts where there are no challenges. Throw in the weirdo ballots with lizard people, stray marks and indecipherable dots, and the error rate grows even more. Throw in the missing ballots, and the situation is hopeless. In truth, the counting errors dwarf the tiny numerical difference in votes between the two candidates. If, at the end of the recount, Mr. Coleman or Mr. Franken is ahead by a few dozen or a few hundred votes, that would be because of errors rather than voter preference.

Seife notes that MN law allows for ties to be decided by lot. I’m not sure if the election board can declare a statistical tie though. Here’s the section of the law I’ve found:

In case of a tie vote for nomination or election to an office, the canvassing board with the responsibility for declaring the results for that office shall determine the tie by lot.

Perhaps it’s too late to do so in this election, but I still think deciding extremely close races by chance could prove more legitimate than endlessly contested recounts.


Recounts of no significance?

It’s been more than a week since the polls closed and we still don’t know who the winner is in three Senate races. Georgia is headed to a runoff. Minnesota is going to have a potentially ugly recount, with only 206 votes out of nearly 3 million separating Coleman and Franken. Alaska is still counting and the difference there could also be small enough to trigger a state-funded recount. Supposedly these new tallies in Minnesota and Alaska will tell us who “really” won the elections. But will they actually mean anything?

Leonard Mlodinow writes about an election strikingly similar to Minnesota’s in his excellent book The Drunkard’s Walk:

In the 2004 governor’s race in the state of Washington, for example, the Democratic candidate was eventually declared the winner although the original tally had the Republican winning by 261 votes out of about 3 million. Since the original vote count was so close, state law required a recount. In that count the Republican won again, but by only 42 votes. It is not known whether anyone thought it was a bad sign that the 219-vote difference between the first and second vote counts was several times larger than the new margin of victory, but the upshot was a third vote count, this one “entirely by hand.” The 42-vote victory amounted to an edge of just 1 vote out of every 70,000 cast, so the hand-counting effort could be compared to asking 42 people to count from 1 to 70,000 and then hoping they averaged less than 1 mistake each. Not surprisingly, the result changed again. This time it favored the Democrat by 10 votes. That number was later changed to 129 when 700 newly discovered “lost votes” were included.

Neither the vote-counting process nor the voting process is perfect… Elections, like all measurements, are imprecise, and so are the recounts, so when elections come out extremely close, perhaps we ought to accept them as is, or flip a coin, rather than conducting recount after recount.

It might be true that, for procedural reasons, later vote counts really are more accurate than the initial one. For example, running the ballots through a counting machine a second time picks up votes that were missed due to infamous hanging chads. In that case a recount could be worthwhile (see a discussion here as it relates to Florida in 2000). Even so, there’s some level at which the difference ceases to tell you anything reliable about who actually received more votes. There’s appeal in this idea of choosing an acceptable level of statistical significance and, when it’s not met, simply letting the original count stand or deciding the election by a random process.

Replacing recounts with with a random selection process would be a tough sell though. One objection is that potential voters may not participate unless we make sure that “every vote counts.” That’s a nice ideal, but we know it’s not achievable in practice. That’s why we have recounts in the first place — we don’t know how to make every count. The best we can hope for is that the recounting process will be more accurate than the initial tally. If we instead accepted the chance that a random event would decide the outcome, people would still have an incentive to vote for their candidate; the baseline would simply be moved from winning by at least one vote to winning by a significant margin. In either case a single vote has a vanishingly small chance of making a difference and voters will presumably be motivated by the same things that motivate them now, such as civic duty, the desire to express their beliefs, or showing support for their party.

Another objection is that deciding a hotly contested election by a coin toss would decrease faith in the democratic process. That might be true, but so do recounts and other tight races. Many Democrats complained throughout the Bush presidency that he stole the 2000 election in Florida and the 2004 election in Ohio. Similarly, Republicans suspected voter fraud long after the resolution of the Washington governor’s race described above. The opposing camps in Minnesota are already gearing up to follow a similar path, with lawyers ready to take their arguments to court when the counting’s through. Regardless of who is declared the winner, we can be sure that the other side will doubt the result and allege wrongdoing. None of this would happen if the winner were decided randomly. He would go to office with the knowledge that he didn’t win a clear mandate from the state’s voters and the voters would know that he at least was given the position through a transparent and fair procedure.

And then there’s the fiscal savings. Minnesota’s Secretary of State Mark Ritchie pegs the cost of the recount at 3 cents per ballot, or a little over $86,000. That’s small change in government terms, but the lawyers’ fees, court challenges, and time consumed add up too. And for what? A result that will likely remain disputed and may not tell us anything about Minnesota voters that we didn’t already know. (Hint: They’re closely divided.) A coin toss costs only a quarter, which can then be given to the loser as a consolation prize.

Regardless, proposals like this aren’t going to gain any traction. That’s partly because people don’t understand statistics, but perhaps more so because they enjoy a good spectacle — and that’s one thing our system of recounts provides with absolute certainty.


No shacks for smokers

Legislators in MN are considering a new provision that would allow bars to build separate “smoke shacks” where smokers can congregate. Employees wouldn’t be allowed to serve drinks in the shacks. Non-smoking customers wouldn’t have to go into them. And given the expense of building a separate structure, its doubtful that all that many bars would choose to build them. Yet even this is not enough to satisfy ban proponents:

Those who opposed a statewide smoking ban can’t seem to take no for an answer when it comes to weakening the ban.

This year, they’re trying again. They want to allow bar owners to build so-called “smoke shacks” outside of bars and restaurants that would allow smokers to presumably stay warm while they light up. Several Democrats and Republicans, including Rep. Bob Gunther, R-Fairmont, presented the smoking shack amendment for amending onto a budget balancing bill shortly before midnight on April 3…

Proponents of the amendment say they want to provide relief to bar owners who’ve been hurt by the smoking ban. But many of those owners already have created patios and other smoking-permitted places for their businesses. Those proponents also should consider the fiscal implications of this exception to the smoking ban. Do the profits of some bar owners outweigh the savings in smoking-related health-care costs to taxpayers?

Somehow I don’t see outdoor patios as terribly appealing places to smoke in the Minnesota winter. The opposition to this extremely reasonable bill nicely illustrates just how much the anti-smoking movement is motivated by the paternalist desire to control others rather than by actual concern for the health of service workers and customers.