A GOOD DEBATE STARTS ON TWITTER
Courtesy of my friend Kurt Peacock, I found myself in a debate on minimum wage and poverty with Stephen Gordon, an economics professor at Laval. Professor Gordon had, as academics should, shared his research with the general public through the Globe And Mail.
The link is worth following, and the gist of the argument is that in Ontario and Quebec following minimum wage increases, the proportion of minimum wage recipients living in poverty was not significantly higher than that of the general population. (Professor Gordon adds additional numbers on how few recipients of minimum wage are subsequently out of poverty after receiving the increase....again, it is worth following his link to read his case as he would state it).
Kurt, whose thoughts I follow with respect and interest, had shared Prof. Gordon's work with a tweet that suggested that New Brunswick's poverty reduction covenant may have been completed without an eye to data. Final forum authors were, in fact, well aware of evidence and debate on both sides of the issue, and shared a link to evidence from Professors Card and Krueger, whose analysis of a minimum wage hike in the U.S. found a ripple effect on wages that did impact those closer to the poverty line. I further submitted that Prof.Gordon had perhaps asked a different question than the one that the authors of the poverty reduction forum asked, as our objectives were about the use of minimum wage policy to allow needed reforms to social assistance and to provide incentives for short-term decisions that lead to employment for those on social assistance.
Kurt linked Prof. Gordon into his reply, and a debate was on. I have provided him with links to the full poverty reduction plan. I think, at some point, he and I have argued at cross-purposes; through no fault of his own, his introduction to the debate has led him to defend his own findings while my argument is that one can be aware of his data and even accept it, but also consider other findings and make reasoned decisions about whether his data tells us much about a new public policy problem.
I decided it was worth a blog post for two reasons. First, I often find Twittter debates turn into a depressing battle of links. Prof. Gordon's data is correct (even if I find his conclusions questionable) and so is the Card/Kreuger data; as with any issue, we can trade these volleys ad infinitum, me raising him a Krugman and he calling with a Friedman, until the debate becomes divorced from the actual problem at hand (helping New Brunswickers in poverty break the cycle and get to decent employment that pays liking wages). Second, it may make for an interesting window for readers into how politicians use data to make decisions.
Solving New Problems
Let's start with an important point - there is no way yet to resolve which one of us is right about New Brunswick's plan. The plan has a commitment to measure and evaluate, so there is a commitment to use evidence. But there will not be a perfect study to predict because New Brunswick had unique challenges and demographics and the plan has never existed anywhere in this exact form. What we do is take conclusions and experiences from other studies and jurisdictions and try to learn what will apply to New Brunswick.
The New Brunswick plan was built around certain premises, which came from engaging people who had escaped poverty and those still in poverty, and from a review of literature to see if our suppositions were valid.
- Our uniquely low rates of social assistance were a barrier to employment, because recipients spent much mental energy seeking food, shelter and emergency funds instead of training and traditions to work. (This was a common finding in reviews of low-rate U.S. jurisdictions)
- Too often, entry level work was a poor incentive for someone to leave social assistance, because it paid poorly and removed some key benefits such as health coverage and child care assistance). So social assistance became a trap, where recipients were paid too little to support the journey to employment and made worse off even if they did make it to a job.
- The way to get those in poverty to the ultimate goal -- stable employment with living wage -- was to make the first step possible. A minimum wage job does not lift one to the final goal, but without that first step people will stay trapped in a social assistance system which is horrible.
-- Minimum wage hikes were not an end in themselves or the only tool, but were needed to solve the underlying problems. If we did not raise the value of minimum wage jobs, social assistance hikes alone would further destroy the benefit of moving from welfare to work, and if we did not raise social assistance rates, they would trap recipients in an endless fight for survival instead of a journey to work.
It is worth noting that this is not a direct challenge to Professor Gordon's ability or conclusions. On most of these questions, he is not claiming to have done research or even to be qualified to do so --others have credentials in this area and we have used their work. It is noting that what he has studied -- whether minimum wage hikes alone have lifted recipients above the poverty line -- is not necessarily the end of the debate about whether minimum wage work is more likely to lead to better employment later, or whether low SA rates help or hurt transitions to employment, or whether an integrated plan like New Brunswick's with all the countervailing tradeoffs will help reduce poverty.
None of this is an attack on my learned colleague's abilities or data. Prof. Gordon deserves credit for adding to the canon of what we know on minimum wage, an area where research has been more focused on testing its impact on employment and less on testing its claimed positive impact. It is sometimes a trap with academics that their expertise in one area leads them to conclude an expertise in areas outside their training, but frankly it seems more likely that the means by which he has been brought into the debate have simply and understandably obscured these other issues (which is why I am writing this, in part, so we can state the full policy question at issue).
An Objection To Gordon's Work - The Trouble With Threshold Evaluations
In explaining why Gordon's work is not as helpful as that of other researchers in the area, like Card, I do have to raise one quarrel with how he uses his data to draw unwarranted conclusions. Astute readers will not an important distinction -- Card et al asks if people's incomes went up, Gordon asks if they were still living in poverty (as defined by the poverty line).
This is a threshold analysis, while Card uses a benchmark test. The difference is important, because they measure something different.
To explain why I think Gordon asks the wrong questions, check out these statements....
Since drafting Alexander Ovechkin, the Washington Capitals have won zero Stanley Cups. Therefore, I conclude that no team should have drafted Ovechkin.
Since completing their university education, neither Kelly Lamrock nor Stephen Gordon have won a Nobel Prize in Economics. Therefore, university education does not make people more informed about economics.
Here you see the problem with threshold analysis -- it may blind us to improvements that show a move was valuable, even if it did not alone take us to a certain level. The Capitals are surely better with Ovechkin than the woeful pre-Ovechkin Capital team. And you don't need to be a Nobel winner to be glad you went to school. A threshold tells us if something alone met an established objective, but it does not tell us if it made us better or worse off. (For more on this, read critiques of the No Child Left Behind education reforms in the U.S., where critics noted schools seemed to put resources disproportionately to the students who were just below the literacy standard, sometimes shortchanging gifted students or severely struggling students who could prove but would not pass the magical threshold when they improved).
The poverty line is just that -- a line of income whereby a family can purchase a defined bundle of goods and services that are deemed to be essential. But what if, en route to having a job that pays at that key level, one first must go through the steps of having higher income in steps that get you within reach of the next step? Gordon's measure misses this entirely. So when he writes, for example, that the impact of minimum wage hikes on other groups would "show up in the poverty data", he is wrong, unless the impact specifically had an impact of taking individuals who start below that particular point to a point just above his chosen line. It would not reflect improvements in income or purchasing power that do exist in very real ways for poor families.
So, I accept Gordon's point that hiking minimum wage to a point still lower than the poverty line does not alone lift recipients over the poverty line. It is undeniably true, but a lot of work to prove a tautology. It is equally tautological to say that 100% of recipients of minimum wage have more money after you hike minimum wage, but that doesn't alone vindicate me. My point is that it is what those improvements allow -- in this case, a less punitive SA rate, more incentives to take entry-level work, and a firmer step toward the middle class for transitional families.
It isn't a critique of Gordon's work to point out that it tells us nothing about the value of New Brunswick's minimum wage hike on the welfare to work journey, just as it isn't a critique of the poverty reduction plan to point out that it does little to renew fish stocks off the coast of Shippagan. Both should only be evaluated based upon what they purport to tell us -- and Gordon's flaw is not that his work doesn't address welfare to work transitions; his flaw would only be if he pretended that it does.
I hope this post has shown some of the challenges decision makers face when trying to solve complex problems using data in areas where we are not, ourselves, academic experts. There is no group called "the experts" any more than there is a group called "the public". After all, Card and Gordon both know more than I do about economic theory, and they disagree with each other. To make things harder, each can add expertise in one area, but poverty is a field which will also involve disciplines where they have no particular expertise beyond being (like the authors of the poverty plan) well-informed laypeople.
Using models from other provinces and states, academic literature on poverty transitions, and learned experiences from those who had overcome and worked with poverty for years, we became convinced that the long journey from welfare to stable employment required SA rates that provided some stability of survival and entry-level work that was a better deal than social assistance, and that these were steps people had to reach to get to the ultimate goal. And that's why Gordon's finding that minimum wage alone doesn't get people to an income beyond the poverty line doesn't really answer the question of whether higher minimum wage with the rest of the policy steps contained will get people closer to being above the poverty line than they would be if we stuck with the status quo.
I was also guided by these questions, which I will ask critics of the minimum wage his in the New Brunswick context.
- If minimum wage does not go up, how would you raise social assistance rates that experts agree are punitive and damaging to the employment transition, and still make the choice of entry-level work over social assistance worthwhile?
- Since 1987, New Brunswick has pursued a unique policy - we have had the lowest social assistance rates and minimum wage rates in Canada. What tangible gains in employment and poverty rates has that policy demonstrated in the last quarter century?
In the end, every policy must start in a place where it has never been done exactly that way in exactly that place. The reason the poverty strategy has a commitment to measure is for just that reason, and I hope it will improve lives here and offer some guidance to those who seek the same in their communities.