The census, evidence-based policy analysis and a reversal of roles

One of the least edifying aspects of the census debacle is the government's spin to the effect that that the only people who oppose its decision to make the long form voluntary are 'left-wingers', so their concerns can therefore be dismissed out of hand. One version of this meme takes the form of the argument that sabotaging the census is part of a broader strategy to diminish the importance of government in the lives of Canadians. If there are no data to guide would-be social engineers, so the reasoning goes, then they will be prevented from expanding the reach of the State into new spheres.

This is a puzzling argument, and not only because it is based on a non sequitur. It betrays a fundamental misreading of the history of the Canadian welfare state and of how evidence-based policy analysis has evolved over the past two generations. Before the census became an issue, the Left, not the Right, was the more determined opponent of evidence-based policy analysis.

The census as we know it is relatively recent phenomenon. Before 1971, governments had access to only fragmentary data sets, and the available resources for analyzing them were, by modern standards, rudimentary. But this lack of information was not an obstacle in constructing the basic infrastructure of the Canadian welfare state: its major features - publicly-funded health care, pensions and unemployment insurance - were all established before the modern census. It’s not clear why anyone would believe that depriving the government of data would prevent it from introducing new programs.

Of course, the lack of data did have important implications for the development of the welfare state: it was clumsily-designed and wasteful. As a result, much of the evidence-based policy analysis that has taken place over the past 40 years consists of documenting cases where existing policies were inefficient, ineffective, or even counter-productive.

But here's the odd thing. Instead of being pleased at being offered the opportunity make the welfare state more effective and efficient, the Canadian Left's reaction to these studies consisted of pretending they didn't exist, or of attacking the authors with such pithy epithets as 'corporate apologist'. There are any number of areas where the same debate played out; here are four.

Free Trade: Those of us of a certain age can remember how confidently the opponents of the FTA asserted that its passage would lead to disaster. Of course disaster didn't happen, and the available data - as analysed in Dan Trefler's famous AER article - suggest that its effects had exactly the effect that its proponents predicted: small and positive.

Pay equity laws: These measures are designed to improve the prospects for women in the labour market, and to close the wage gap with men. It turns out that the experiment in Ontario in the early 1990's was largely unsuccessful in attaining these goals.

Employment Insurance. A poorly-designed employment insurance program can be abused, and as I noted in this post, that's exactly what happened before the reforms of the 1990s. To those studies, we can add this one by my former grad school office-mate Jane Friesen, which documents how those reforms produced the predicted behavioural responses.

Social assistance. Economic theory predicts that benefits that are too generous provide a disincentive to find employment, and available data are consistent with the theory.

It would be a stretch to characterise these research programs as being part of a big-government plot, and they were invariably denounced by conventional leftists. Moreover, these denunciations were almost never evidence-based. Evidence-based criticism takes the form of "The study doesn't take X into account. I should incorporate X into the analysis and see if the results are materially affected." Instead, critics would content themselves with "The study doesn't take X into account, so I'm going to squat stubbornly on my original position." In retrospect, trashing the census might be something that you'd expect from a NDP government: as far as they would be concerned, evidence-based policy analysis has been a source of embarrassment.

Of course, none of this lets the Fraser Institute off the hook. Those of you on the FI's e-mail list (I'm pretty sure that usually includes me, but they seem to have forgotten this time) may recall receiving a fund-raising thingy last week; it is reproduced here. If you click on the link to 'academic elites', you get sent to this post of mine here on WCI. Now, there are two things you should know:

1. As anyone who *is* part of the economics academic elite will tell you, I'm not one of them. My CV allows me to claim 'respectable' status, but no more.
2. My criticism of the Fraser Institute's position is not based on its stated horror of state coercion: de gustibus non est disputandem. What the FI got wrong - unforgivably wrong for an organisation that claims expertise in policy analysis - is the claim that the census decision would have no effect on the quality of the data and hence of the ability of citizens to hold the government accountable for its actions. This is not a matter of opinion; it's a matter of mathematics, data and science. The fact that the Fraser Institute doesn't acknowledge this - and chooses to distort the point its fund-raising campaign - demonstrates that it has no interest in evidence-based policy analysis.

The census decision has been denounced by the evidence-based policy analysis community, but the real test of membership in this group is not opposition to the Harper government's position. It is a willingness to let their policy views be guided by the evidence, and to modify those views if the facts change.

Comments

[Mark]

I would say "The study doesn't take X into account, so I'm going to squat stubbornly on my original position," is still moderately respectful of evidence-based analysis, unless X = "the power of magical thinking." Most people lack either the time or the ability to reproduce the work incorporating some new consideration. So yes it may be grasping at straws to maintain their worldview, but it's at least doing so within the framework of evidence-based analysis. That's not the same as saying, "We don't need no evidence-based analysis, and I will suggest nothing that might refute your analysis because my gut says you're wrong."

[Stephen Gordon]

Perhaps, but that's actually a fairly rare position; I was being generous. Deferring to one's gut and/or shooting the messenger is more prevalent.

[Stephen Gordon]

And unless X is a fatal flaw, the data should still force a partial revision in their position.

[Jim Rootham]

Some of us think that becoming a a US style plutocracy IS an economic disaster and the FTA was a significant component in that change. Less because of its trade effects and more because of a general tilt towards unrestrained corporate power.

[Mike Moffatt]

This is the best/most important blog post in the history of WCI (and please don't take that in a 'damning in faint praise' way!). I hope this gets heavily cited and republished.

[Jim Rootham]

Also, on the last two, some of us think those are acceptable/desirable results.

[Mike Moffatt]

"Less because of its trade effects and more because of a general tilt towards unrestrained corporate power."

I would love to see a study on whether 'corporate power' has increased in Canada over the past 200 years. I have trouble believing this. Go back to 1950, where a couple of dozen Anglo families in Toronto and Montreal practically owned all of Canada (the Bassetts, the Eatons, etc.) Is political and economic power really more concentrated now than it was, say, 60 years ago? Or, say, in the heyday of the Family Compact?

[Frances Woolley]

Steve:

1. Pay equity doesn't address employment opportunities. That's employment equity.

2. There are many different types of pay equity laws (equal pay for equal work, equal pay for work of equal value, pro-active legislation, complaints based legislation, etc).

It would be dangerous to generalize too much from one particular study of one particular form of pay equity legislation.

Possibly pay equity is legislation in Ontario didn't have much of an additional effect over and above the federal and provincial human rights legislation preventing discrimination and federal pay equity legislation that was already in place. Possibly the implementation of the legislation left something to be desired.

This says nothing about the effect of introducing pay equity legislation in a context when equal pay for equal work is not already guaranteed.

[Christopher Kam]

All the arguments about evidence-based policy, the statistical validity of a mandatory versus the invalidity of a voluntary are quite secondary to a far more important issue: Do you think it appropriate that the Canadian government (or the government of any half-way liberal democratic society) be able to jail a citizen for up to 3 months for refusing to answer questions about, e.g., how much unpaid housework they do in a month? I, for one, think it highly inappropriate; only an authoritarian (and you find those on the left and right) would argue otherwise.

I should note also that I put little stock in the counter-argument that the government has not tended to resort to jail time to enforce compliance with the census. In the first place, it is disturbing enough that they could. In the second, if jail time is an empty threat, then remove the law from the books: nobody is made better off by keeping a fictional statute on the books.

[Stephen Gordon]

You may have missed it in the news, but the government is removing the threat of jail. No-one seems ready to vote against the measure. Problem solved.

[Mike Moffatt]

Also keep in mind the jail provisions are pretty standard in *every* Canadian act. Under the new 'Canada Consumer Product Safety Act' you can (at least, in theory) be jailed for *years* by using the wrong font size on a product label.

Do you think the Conservatives adopting that measure is appropriate, Christopher?

[Jim Rootham]

@Mike

If you want to say that historically things have been good for the powerful and bad for the powerless I'm not going to argue.

There was a short period immediately after WWII when things looked like they where going to get better for the bottom of the range. I am looking at income distribution in particular. That pretty much stopped happening immediately for the very bottom, and over time the lack of improvement and decline reached higher and higher into the income distribution. In the 70's(?) that process reached the median income earners and somewhat after that reached the average income earners. The process has accelerated since.

[Mike Moffatt]

"In the 70's(?) that process reached the median income earners and somewhat after that reached the average income earners. The process has accelerated since."

Agreed - no argument there. The change, however, has been in labour income, not capital income so I'm not sure if it tells us much about 'corporate power'.

[Kevin Milligan]

Frances: Stephen doesn't say that pay/employment equity laws are bad. He states the finding: "It turns out that the experiment in Ontario in the early 1990's was largely unsuccessful in attaining these goals." That is indeed just one study on one aspect of the broader policy. But it is not a finding supportive of that kind of law.

But, I believe the point Stephen is making here is that this study clearly didn't advance the "socialist" agenda, which is what is alleged by census critics about users of data.

The 'data users are socialist planners leftists' thing is just blatantly in direct opposition to my experience over the last 15 years in this biz.

[Kevin Milligan]

About free trade. Stephen's point is that pointy-headed academics used data and empirical evidence to argue in favour of a free trade agreement, which was a policy advocated by more conservative governments. This is the opposite of the 'data users are socialists meme.' You don't have to like free trade or 'global corporate hegemony' to accept that point.

[Christopher Kam]

Steven, excellent: with the jail provisions removed the long-form is, in essence, voluntary -- and this by all-party unanimity. To wish that they would also scrap or even just reduce the fine to a nominal amount! (An aside: I find it sad and ironic that we are subject to a fine for failing to answer census questions but not for failing to vote despite the fact that the latter is a far more integral aspect of democratic citizenship. And - NO! - that's not an argument for fining people for not voting; exactly the opposite.)

Mike, the fact that so many of our statutes include provisions for jail time is troubling with a capital T. This is not a matter to be cavalier about. First, jail should be a last resort to deal with the violent and incorrigible. Second, I suspect/faintly hope that many of these provisions are dead letters, but if so and as my initial post noted, having a quasi-fictional statue book is to nobody's advantage.

[Mike Moffatt]

Christopher: Agreed - pretty much all of these provisions are "dead letters" as you describe. I'm good with getting rid of them as well. It was a little ironic to hear Tory MPs cry out against the jail provisions of the Statistics Act WHILE AT THE VERY SAME TIME putting even harsher provisions in the Canada Consumer Product Safety Act.

[myron]

"Also keep in mind the jail provisions are pretty standard in *every* Canadian act."

There are good reasons to have the discretion to have jail time at the far end of the sanctioning process in these statutes. I don't know that "dead letters" is a fair characterization. Toothless laws with no viable means of enforcement would be worse, IMHO, than statutes with punishments that are rarely or never used. At the end of the day there has to be consequences for flouting the law. And I don't think anyone is cavalier about jail time. Look at all the bloviating Cons will do about light sentences for criminal activity. Judges are not eager to gravitate to jail time, and there are legal principles that that approach is based on. But it needs to be there for the rarest of rare instances of the individual that has no intention of ever complying.

[Frances Woolley]

"The 'data users are socialist planners leftists' thing is just blatantly in direct opposition to my experience over the last 15 years in this biz."

It all basically comes down to one policy issue, which I've talked about before in the comments before: employment equity.

Employment equity basically requires employers to identify and remove obstacles to the employment of everybody except for able-bodied white males (i.e. identify and remove obstacles to the employment of people with disabilities, members of visible minorities, aboriginal Canadians and women).

Now since able-bodied white males are more likely than just about any other demographic in Canada to vote conservative, you can imagine that this policy doesn't exactly go down a storm with the conservative party base. Plus it's much easier to blame the fact that you can't get a job on your skin colour than your competence.

What does this have to do with the census? Suppose you're a business looking to hire someone with a PhD in economics. In order to know whether or not you're achieving your employment equity goals, you need to know how many aboriginal Canadians, women, people with disabilities and members of a visible minority have a PhD in economics. If your organization has 100 economists and not one is an aboriginal Canadian, a good way to justify that would be to say "there are only five PhDs in economics in Canada who are aboriginal Canadians, we're just not going to be able to find an aboriginal Canadian who's qualified to do this job."

There is no data base other than the census that provides you with that information.

Kill the census, you kill employment equity.

Now, you could just kill employment equity without killing the census, and it seems that this is likely to happen in any event, but killing the census would be the stake in the heart, preventing employment equity from arising, vampire like, from the dead....

I agree that what Stephen said was not, strictly speaking, inaccurate, but it could easily be misinterpreted.

[Jim Rootham]

I get the main point, and I agree with it. I will plead guilty to thread drift, but I am going to do it anyway.

@Mike

I have a theory that says the increase in labour income at the top of the scale is actually returns on capital captured by the managers of large firms. Or more generally, managers are charging economic rents. Who is paying the rent is less clear. Some is clearly being paid by people lower in the income scale, whether any is being paid by shareholders is not so obvious.

This theory should be tested before doing anything really substantial with it, but the thirty year disconnect between productivity and median wages is pretty good evidence to me that something is broken (even if broken is the historical norm).

[Mike Moffatt]

"I have a theory that says the increase in labour income at the top of the scale is actually returns on capital captured by the managers of large firms. Or more generally, managers are charging economic rents. Who is paying the rent is less clear."

I really, really wish we had a better idea/data on who the top 0.1% income earners are and where they worked. Then we would have a better idea of where the money is coming from. I suspect many of them are professional athletes and small/medium sized business owners, but who knows?

In the professional athlete case, a lot of it is being made up by owners. Back around 1960 the Dodgers were earning over 4 million dollars a year in profit, while paying Koufax and Drysdale 20-25K each despite the fact they pretty much were the team. Capital was able to 'exploit' the workers thanks to a cartel arrangement and the reserve clause - a situation which was greatly lessened by Marvin Miller and the Players Association in baseball, and the formation of the WHA in hockey.

In that case, there is a big disconnect between athlete salaries now and athlete salaries 40 years ago, but that's only because owners were robbing the players blind back then.

[Matthew]

"Economic theory predicts that benefits that are too generous provide a disincentive to find employment, and available data are consistent with the theory."

Stephen, how do you square this statement with you advocacy of a Guaranteed Annual Income? I ask because I have also spoken in favour of this to friends and colleagues who dismiss it for this very reason.

[Stephen Gordon]

The incentive structure is different. Conventional welfare penalises those who get a job; a GAI doesn't. And there are risks if the GAI is too generous, too.

[20th century workforce]

Maybe grandfather existing welfare, disability, EI, and pension roles, exempt them from GAI, start out GAI at $5000 and add a negative income tax that tops out at another $5000? Social workers would make good nurses and employment counsellors.

[Jim Rootham]

Of all the big pay cheques to worry about, professional athletes are at the bottom of the list. There aren't enough of them to have large economic impacts, they have relatively low political impacts (not zero, Jim Bunning is a significant negative political impact).

We can get a handle on how many big earners are athletes. There are 6 NHL teams, the NHL minimum makes it about 20 per team in the to 0.1%, so 120, the Raptors, 15 more, the Jays, 20 more, 8 CFL teams maybe 10 per team, so 80, skaters, golfers and tennis players etc. another 20?

120 + 15 + 20 + 80 + 20 = 255

Almost certainly less than 300.

There are ~20,000 earners in the top 0.1% so it sure looks like athletes don't have much impact there.

OTOH a big fraction of that 300 will be in the top 2,000 so at the very top it is significant.

One trick here is the wage structure of athletics, it's extremely winner take all.
That will also apply to entertainers generally. I can't see any more musicians+actors etc making that kind of money, probably less.

The open question is small/medium business owners vs large corporate managers.

I'd bet on the managers.

[Mike Moffatt]

Jim: You may be right - it's unfortunate that there's no way to tell.

[Jim Rootham]

You ought to be able to get into the ballpark. We have the top salaries at public corporations data. We have salary survey data. We can subtract out the athletes. It should be possible to build a model of income distribution in large firms with a parameter that describes the steepness of the curve. We know how much money we need to account for. That should give a range for the parameter. The result ought to tell us something about who is getting the money.

[Just visiting from Macleans]

Who is paying the rent is less clear. Some is clearly being paid by people lower in the income scale, whether any is being paid by shareholders is not so obvious.

Take, for example, stock options - it's principally the stockholders who pay - one reason why Warren Buffett had long lobbied to have them included in financial reports.

A Canadian example - when Jim Balsille and Mike Lazaridis(co CEOs of RIM and significant RIM shareholders) "negotiated" to pay back something like $75 million of compensation improperly obtained by pre-dating their options. As managers/owners in that industry - they know the key to their success is hiring and retaining top people - so they're not going to cheat their employees out of funds that they would have been entitled to - it was the other shareholders who were stiffed/unstiffed.

[Jim Rootham]

Stock options require a whole analysis unto themselves. Certainly there is a prima facie case to be made that stock holders pay, via dilution if not by transfer for them.

How that factors into pay scales down the chain is not so obvious.

The process by which brass collects economic rents is much more ubiquitous and insidious than that. RIM pays competitive wages, so they are part of any stiffing that is going on. Note that the stiffing gets worse the lower you are on the pay scale, it's the janitors who are really getting it in the neck more than the programmers.

[Just visiting from Macleans]

But, if you're suggesting looking at the top brass pay compensation, you have to take stock options into consideration - for many it is the dominant form of compensation.

You've got to have a whole batallion of janitors to stiff them collectively of $75 mill; and this is no longer like the era of Mad Men with whole banks of receptionists, secretaries, typists. I doubt there's many at RIM who are not carefully recruited and retained. And probably participate in profit sharing (another animal) and perhaps stock options well into the lower levels. The rules for companies that compete on price (which is not the segments RIM operates in) don't apply as well. But mostly O/T, I'll move on.

[Karen Krisfalusi]

We know that the census, mandatory or not, is a grand survey with questions whose answers will guide policy -- policy that is already framed to a particular purpose. If the government is not interested in writing a survey whose results will educate and inform 'policy' (far moreso than create new knowledge) isn't this the key 'role reversal'?

[Jim Rootham]

Stephen, I was just catching up with reading Bread and Roses, and I just realized you may be talking about your experiences there and at Babble. Are you?

It took a while for the penny to drop because your description of what happened in your post here is so wrong as to be unrecognizable.

First let me point out that the left in Canada is the antithesis of monolithic and there are almost certainly some twits who responded to your posts in exactly the way that you describe.

I observed (and was part of) the opposition to lot of your recommendations. I will admit that some of the motive imputation on my and others parts was less than useful.

However, most of the substantive opposition to your recommendations was not based on rejecting the data, or rejecting the idea of getting good data. It was based on a different value system that said the questions being addressed by the data were irrelevant, or not sufficient reason to change the choice, or in some cases actually a desirable outcome.

As an example of the latter, you say that generous social benefits reduce the willingness to work, or in economic model terms, reduces the elasticity of the labour supply. You have also pointed elsewhere out that a lower elasticity of labour would imply that more of the tax burden is carried by owners of capital. I am also assuming that you think this will reduce the efficiency of the economic system (otherwise why would it be a bad thing?). The conflict between you and me over whether this is a good thing has nothing to do with idea of using evidence based decision making. We both think that is a good thing. The conflict is about what constitutes a good result.

[Appalled but not confused]

I think a number of things are being conflated together in Prof. Gordon's post which shouldn't be, given the context of the debate.

Some of your points here were well & rightly made & should be well taken, to be sure (even if some of them were against straw men):

- that the Cons. are wrong in clearly wrong in claiming that the only user groups complaining are those on "the Left" &/or whose analysis would only justify increased govt spending or programs;
- that not all the gov't programs _were_ established on the basis of evidence, census or otherwise; and
- that "The Left" has always been eager to get or receptive to evidence-based arguments.

But none of that vitiates the claim I made here on July 2,* or that Taylor & many others have made since, that one of the primary motivations behind Harper's decision on this was likely to ruin the quality of data on vulnerable groups that _has_ been used for the past 20 years or so by advocacy groups like Social Planning Councils and The Cndn Center For Policy Alternatives in things like the 'Campaign 2000' anti-child poverty initiative, and by womens' & ethnic & immigration groups to track and try to improve their economic progress, and so on.

Most of these groups don't do much in the way of analysis: they just want to confront the gov't with the broad indicators, that x% of y groups are still living under the LICO poverty line; that the levels of educational attainment & income of y group is less than z; and that / whether single parents in provinces w/o subsidized daycare have lower income, less employment, & worse occupations; and so on.

And with all those indicators coming from the long form census & the 2011 series being ruined particularly for the populations they're most interested in, those groups won't be able to make an effective case that things have gotten worse when they try to advocate for more subsidized daycare, ESL, employment training etc. etc. -- which is probably why Harper did it, because he's demonstrably hostile to these "interest groups" demanding more money to advance such programs.

Admittedly I'm a little under the weather right now, but I don't see anything in the post that contradicts that.

* http://worthwhile.typepad.com/worthwhile_canadian_initi/2010/06/an-incredibly-stupid-decision-on-the-2011-census.html

[Steve Thompson]

Being part of the Labour Force Survey for the next 6 months, I have learned something about Statistics Canada's methodology. Their initial questions are labour-related; how many hours did you work, were you looking for work etc. Then, they throw in the zinger. This month, the question was about trips over 40 kilometres that you've taken over the past month (i.e. July). They wanted to know how many trips were taken, the purpose of each trip (medical, business, work, social etc), how much was spent on food and beverages, how much was spent at a grocery store, whether or not you used public transit, how much you spent on cultural events, how much you spent on sporting goods or events, whether or not you purchased season's passes, how much you spent on gas for each trip. The questions took about 15 minutes to answer. Since we live in a rural area, any trip that we take to one of the nearest cities is over 40 kilometres one way and we make roughly 10 of those a month. As well, we were asked what our household income was. Certainly, if the government really wants or needs accurate household income data, it has internal sources (i.e. CRA) that can more accurately answer household income questions than Statscan.

Statscan claims that they have no way to get the data that will now be lost with the changes to the long form. I suggest that they have already found the replacement. Apparently, as part of the Labour Force Survey, Statscan has the right to ask you questions about "other topics of interest". Statscan also touts the accuracy of their statistical data. I would argue that asking me (or anyone else) about trips that were taken over a month ago when I haven't been given a chance to prepare my answer (or record my trips) makes the accuracy of the data collected highly suspect at the very least and probably statistically inaccurate at the worst since you are not granted the option to not answer the question if you don't know the correct answer.

http://viableopposition.blogspot.com/

[Stephen Gordon]

The LFS uses census data to make sure its rotating panel is representative of the general population.

[Mandos]

I must echo Jim Rootham's criticisms, being one Stephen's regular interlocutors at B&R and at babble: the relationship between academic economists and the left is poor not because the left rejects the results of economic work, but because economists tend to ask these questions with implicit values in mind that are not really shared by the left, and working assumptions that the left rejects almost by definition. My discussions with Stephen demonstrated this over and over.

Left-wing economic thought is not merely a naive income- or wealth-levelling exercise. Almost all of the negative political consequences of free trade have come to pass. Universal post-secondary education is a positive goal because we want to break the *political* and *cultural* idea that education is something other than a fundamental human right. And so on and so forth. The idea that wealth is relative---important to the left---is simply unaddressed by assumptions that involve Pareto-anything. These issues are rarely addressed by economists until the negative consequences have *already* come to pass.

For example, apparently we are in a world in which mercantilism works.