The release of two new dueling studies of Seattle’s experience with its minimum wage ordinance has unleashed a flurry of praise and pushback from observers in this much debated field. That they followed closely on the heels of Ontario’s announced plans to push toward a $15/hour minimum wage by January 1, 2019 only reinforces the need to ensure that social engineering policies be based on or, at the very least, be accompanied by adequate analysis.
By now, pretty much everyone in the developed world knows that the Seattle studies came to opposite conclusions. The first, conducted at UC Berkeley, concluded that, so far, there have been no adverse employment effects from Seattle’s gradual raising of minimum wages from $9.47 to $15 per hour within city boundaries. The second study from the University of Washington (UW), using different data, concluded that the negative effects were slight at first, but once the wage floor hit $11/hour, it started to curtail the number of hours available—and not by just a bit.
For those only relying on others’ interpretation of the results, they can be forgiven for thinking this is just another case of economists disagreeing with each other—something they are (in)famous for. The political implication, unfortunately, is that in the absence of research unanimity, the policy that ‘feels right’ is the best one to follow.
For those who have actually read both studies, however, it is hard to give their findings equal weight (there are lots of good neutral comparisons like this one for the less technically inclined). Economic linkages are rarely simple and good datasets are exceedingly tough to assemble, but Berkeley’s analysis is far underpowered compared to UW’s. The two studies don’t even appear to be independent, because there are observations that the Mayor of Seattle—hardly an impartial observer—quickly commissioned the Berkeley study to blunt the effect of the much more comprehensive UW analysis, which must have been presented earlier in draft form.
Berkeley used off-the-shelf employment and payroll data for the restaurant industry from the Bureau of Labor Statistics. It, like most data sets, is wide, but not very deep. The definition of employment is a stark binary condition: either one is employed or not. It doesn’t matter how many hours a week one works or their employment status (permanent, temporary, etc.)—everyone is put in the same group. Payroll data similarly reflect total wages across entire employee groups, mixing full-time, part-time, high-, low-, and middle-income earners. Relying on gross payroll divided by gross employment to arrive at a proxy average wage is the economics equivalent of using buckets for chemistry experiments.
With such blunt data, it is no surprise the Berkeley authors say they could not find any precise negative employment response to higher minimum wages—the data are far too aggregated and homogenized to be sensitive to any single influence. The disappointment is in how categorical they are in their conclusions, since their study did not use direct measures of wages or identify low-paid workers.
The UW, study, on the other hand, goes to great length to discuss the limitations of the data that many previous studies have relied on—including those used by Berkeley’s. And, more importantly, it lists out the caveats to their own research—a good sign of study neutrality. The authors point out that they neither sought nor received study input from labour or business groups.
Unlike Berkeley’s, UW’s data set included far more granular data, identifying actual wage levels and hours of work for individual workers and business units. It also used data for all industries, not just the restaurant sector.
UW’s conclusions are pretty clear: the Seattle ordinance indeed reduced employment underneath the new wage floor in Seattle compared to neighbouring municipalities (as expected, since such pay levels are now outlawed). However, the study did not find any significant relative hike in employment at pay levels at or just above the minimum wage. Because raw employment (à la Berkeley) showed less movement, the implication is that low-wage employees lost hours, not jobs.
The analysis focussed on single-establishment locations because multi-establishment businesses did not submit detailed payroll data separately for each of their locations—a point seized upon by some commentators claiming the study was not representative. However, UW checked separately with multi-locations businesses by survey and indeed found negative employment effects.
Those findings make perfect sense in the real world. A small business with four employees may not drop a full staff position from the payroll because of a modest minimum wage increase. Instead, they would likely try to allocate hours more carefully. For larger enterprises operating across ordinance boundaries, they could have the additional ability and incentive to shift hours or jobs from Seattle to neighbouring municipalities.
The UW study goes so far as to suggest that low-wage employees as a group are worse off than before because the hours they lose more than offset the gain in hourly wages. In net terms, the average low-income employee is $125/month worse off than before the minimum wage increase. It remains to be seen whether that finding holds up over the longer term, or is generalizable to other jurisdictions. UW promises to investigate further in the future and extend its analysis using longitudinal employee data—the gold standard of research data sets.
The study of minimum wages in Canada doesn’t suffer as many data limitations as those sourced in the US, where wage ordinances are municipally driven, or state-by-state differentials are relatively minor. Here, province-wide changes in minimum wages, some of which have been quite large, offer more opportunity to obtain statistically significant results.
The consensus effect in Canada, even using fairly basic (bucket) data is that a 10 per cent increase in minimum wages has been associated with a decline in youth employment of 3 to 6 per cent—and to a somewhat lesser degree to adult workers. If one had more detailed hourly data like UW was able to use, or access to longitudinal tracking information on an employee-by-employee basis, the evidence would likely be even stronger.
Instead of a quest for insight, though, the Alberta and Ontario governments appear to show no interest in actually measuring the impacts of their actions. A cynic might believe that these governments think they have little to gain from doing so. Rather than fearing punishment for driving up youth unemployment or making it more difficult for new Canadians to enter the workforce, they instead see it as a political opportunity to ride to the rescue of these groups later with even more new policies, programs and interventions.
For now, provincial governments seem to be swayed more by simple letters containing the signatures of 50 labour-backed economists than with actual research findings. That stance is a shame. The size of the minimum wage increases planned by both Ontario and Alberta are far larger and quicker than any policy in place in the States. Their impacts are likely to be felt quite soon.
While the focus and noise from labour groups is on challenging the effect of minimum wages on employment, there is almost complete absence of clear analysis from them on their effect on poverty—the erstwhile justification for wage floors.
Although researchers may relish a golden opportunity to measure these effects, the fact remains that the policy shift is still a piece of dramatic social surgery on individual employees and employers, with no anesthetic. Provincial governments are imposing significant changes to the relative cost structures of business. Wetaskiwin, Alberta and Brockville, Ontario are not Seattle, yet they are in line for similar treatment. The averages of typical economic datasets, where even available, will most certainly cloak reality and the ricochets will certainly affect some business owners and employees much more than others. Some may benefit, but others will almost certainly be worse off.
Good or bad, it is not asking too much for governments to care about the true results of their actions.
Ted Mallett, Vice-president and Chief Economist