Farewell Ivan Fellegi

I was recently appointed to the National Statistics Advisory Council and, in that capacity, was present on Thursday and Friday at the last Council meeting for which Ivan Fellegi will hold the post of Chief Statistician. Having served more than 20 years in that post, Fellegi is something of a legend in a town where the typical Deputy Minister serves far, far shorter terms, often little longer than a year or two.

Fellegi has profoundly shaped Statistics Canada to the point where it is currently ranked by most experts as the best in the world. Under his tenure, the agency has produced not just the (exceptionally reliable) numbers we need to conduct basic social and economic analysis, but  also many pioneering special surveys, and a raft of analytical studies which have framed Canadians understanding of the data. Over many years, many of us have looked to StatsCan for well-grounded and informative studies on what we on the progressive left see as the key issues – most notably trends in earnings and family income inequality, and the extent to which some Canadians are being left behind even as national income grows.

The National Post has charged StatsCan with promoting class war for the way in which it packaged the release of key income data from the 2006 Census – and the release indeed appropriately highlighted the growth of earnings inequality, gender inequality, and the lagging fortunes of recent immigrants. The publicity given to the inequality story was the result, in significant part, of hard advance work on the part of StatsCan itself.  We are all the better for it.
I’ve sometimes criticized StatsCan in the past for not being too forthright in drawing policy-relevant conclusions from its analytical studies. The agency rightly seeks to be, and not just to be seen to be, “neutral” – not least because there are enormous pressures from the government side not to highlight some of the more uncomfortable facts of our national condition.

To my mind, and that of many others, Ivan Fellegi made a huge contribution by striking the right balance and thus preserving the independence of StatsCan,  and by building up on organization of impressive capacity.  A recent public opinion survey found it  is one of the most highly trusted federal institutions from the perspective of ordinary Canadians.That’s no mean accomplishment in today’s world.

It is to be hoped that the Fellegi tradition will endure as he transitions into new roles (nobody expects Ivan to actually retire!)


  • Andrew, while I appreciate the points you make and found them to be quite insightful, I’d quibble a bit with a completely uncritical review of Felligi’s tenure.

    First of all, Fellegi is often quoted for his stark statements that the Low Income Cut-Off is not a poverty line, yet has refused to develop an actual poverty line. In fact, the most interesting addition to poverty data was the Market-Basket Measure and that was developed by HRDC.

    Second, under Fellegi’s tenure we have seen increasing “cost recovery” for data that should be in the public domain. This is data that is collected for free from the volunteer time of thousands and thousands of Canadians, and that is financed by all Canadians through their taxes. To not put all data in the public domain for free (as the US does) prevents all but those organizations with substantial financial means from getting access.

    Finally, notwithstanding the great work done by Garnet Picot’s research group, Statscan overall has needed a lot of prodding by groups like the CCPA to do some really important data collection. The wealth survey is case in point. After at 15 year hiatus (the US does it every three years) Statscan finally did a new wealth survey in 1999, but it took Steve Kerstetter, working for the CCPA, to get special runs commissioned and that were subsequently published by CCPA. At least there was another update in 2005, but the future of this important barometer of inequality is very much in doubt.

  • One could also add the “class war” Statcan has been pursuing against its own employees.

    Take a look into how the front line Stat Can interviewers across the country are treated (the ones who actually collect the stats being praised). This is a part-time, feminized work force that is treated as expendable (I know I used to be one). The “sweat shop” conditions found among contract workers in the last census is only the tip of the ice berg.

    While the union has made many gains (since the 1995-era slashing that created the mess and in the face of threats of contracting-out), I wouldn’t be surprised if you found frontline Stat Can employees are among the lowest paid in the federal civil service.

  • Ivan, is, you know, just his guy. A bit of an odd sort. Campy in his looks, and a bit overly optimistic in his visions. I could mention quite a few things in a negative capacity. So lets push the dynamics of blogs and see what happens. I am game, the way I have been abused by someplace I have nothing to lose. Injured but not forgotten.

    I think the one that stands out to me the most is his complete misreading of the environment and the lack of data being collected on it. In the year 2002, the entire budget for the environment statistics programs was more than measly pathetic. (less than $5m) Oh that is so being caught with your pants down. There has been some improvement but not much.

    I will admit that there are some real bright spots to his work, like letting the aforementioned Picot led shops do their work. Garnet was and is a great man and strove to make sure balance was within the sights and Fellegi did well to ensure these balanced voices stayed vibrant and alive. I think their voices have been muffled a bit since Harper came around.

    I also am not sure about how Statcan rates internationally. I do recall that if statcan says it enough, like they do, than enough people will start believing it. Statistical reliability is all in the eye of the beholder, it is something that can shake the foundations, if one wanted to be critically direct. I can say this, Ivan is the father of one of the best data imputation algorithms in the world, (he invented it and it pervades the organization) but when it is abused one of the worst solutions to a poorly responded survey program. But hey I don’t want to shake the foundations.

    All in all we could have done worse for sure. I think it was a mistake to let somebody stay in the place that long. Come on, Andrew is this a democracy we live in or a thinly veiled socially constructed notion of one. The latter no doubt and having one man in charge of the stats barn for that long is pathetic.

    Numbers within this age of reason and scientific measurably constructed notions of everything and anything and the reins over those horses is way too much power to be treated as some kind of joke that we have let one guy run the barn for this long. New blood, new vision, change, innovation, insight. These are words that have a quite hollow ring within that barn. They should have a limit, or upper bouind. How do we know Felligi did a good job, we have nothing to compare his actions to. No objectivity. Sure we can attempt international comparisons. I am not sure how subjective those experts really are.

    For well functioning survey programs I would look south of the border to the BLS. They smoke us on data quality and nice healthy samples and money spent. Okay so maybe the balance issue is not quite what we have, but we are not much further walking on an even beam.

    Yes Felligi is a nice man, but to me he just hung around for way too long. It is not what a democracy should have. And we all know that his arm’s length status afforded to him by the Statistics Act gave him a lot more potency than any Deputy Minister. Don’t you think this to be quite odd really. To me it is just another sign of our backwards, bush league, elite driven institutional controls over this country that otherwise has a lot of potential. lol.

    Sorry Andrew I have to give you a thumbs down just like I gave Bob Baldwin a few years back in his assessment.

    I mean listen to this one, the Workplace Injuries Statistics were allowed to be pulled from statcan and handed over to the association of provincial compensation boards back in ’98. I don’t think Ivan even understood what was up with that. It was perceived just to be a cost saving measure internal to statcan. (I knew differently but was just a peon).If he did have some notion of balance then I am sure they would have fought a lot harder to maintain control over the Injury file. Having the WCB associated boards keep their own stats is like handing the keys to the chicken coop to the fox. Its no wonder those boards and worker comp programs stink so bad these days. No accountability. I would really like to ask Ivan what he was thinking on that file turnover, more like a fumble in your own end zone.

    Similar to the EI program, one of the important reasons to keep that file at Statcan, is to maintain some external validity to the HRSDC administrators. I can say this when the GST file was transferred over to statcan, I am sure the GST administrators just about had a bird. I am sure the entire file was never transferred. External control; and departmental accountability is a huge issue within government and data is the smoking gun.

    Statcan plays more of a role than just counting beans, but these are whole different stories and I am not sure how tuned to those circles Ivan really was.

    My point for raising this, is seeing that April 28th just passed, I am hoping that we can one day see the Injury file transferred back to statcan.

    Just a small add on to this long discussion.

    Okay that is a big rant for the day and will i have a future. Not sure I have one now anyway. Oh well,

  • Andrew Jackson

    I guess I could have been more critical, and some of the points made in comments are valid. However, I do know first hand from OECD events that StatsCan has a pretty high international reputation, and you can’t really blame the Chief Statistician for lack of funding for key surveys – such as on wealth and thre workplace and employee survey – which have been highly irregular or dropped. A lot of the add on surveys to the basic line up in recent years have been funded by federal departments rather than out of the shrinking StatsCan core budget – with HRSDC and the Policy Research Initiative contributing major shares of the overall budget to get what they want.

    On the poverty issue, I used to think as you do Marc, but on reflection I think an official poverty line should be defined politically rather than by a statistical agency – though I also think LICO is a good approximation. I don’t think StatsCan would object to regularly reporting on a series of measures – including the MBM – but HRSDC owns and controls the MBM.

  • Point taken on the Chief cannot ensure funding levels.

    However there are measures that the chief can enact under the Statistics Act that will ensure that the political winds determining what surveys get funding and which don’t are minimized. That is a fact and one that I am sure Mr. F. could have used say for the environment.

    I also believe there has been an abandoning of any worker centric surveys. For example even the Workplace Employee Survey ultimately became a glorified productivity survey. It had a lot of potential to tap into workplace specific worker issues and it did not. It didn’t even have the bare minimum in covering off worker information needs like number of work injuries at a worksite. Could have easily nailed that labour motherhood issue but alas, too much of a anti worker wind blows at statcan. Not sure what post-Fellegi will look like, but could we at least have a shortening of the term.

    We do not have one occupationally based survey at Statcan. That is one survey that defines occupation and samples on it. We have the census long form but even that is a 20% sample. I am amazed how mass production era these statistical programs are caught in. Why, I wonder, maybe the reign was too long. He ruled with his ear real close to the ground and definitely had his say on many of the survey programs developed.

    How can it be that in this age where everybody preaches that the most important resource is a human resource, we do not have one survey that focuses on the occupationally dimension in a manner that provides sample controls for that variable. You would think that having at least a count of the countries Human Resources would be a lot higher up on the list of priorities. With all the education that goes on in this country it would make a lot of sense to see how these early assets are implemented and trasnformed within the labour force. The LFS has an occupational component but because of the sample controls do not take into account occupation, the results cannot be used at any level of detail that would promote even a general understanding of occupational detail.

    This too me is a sheer lack of insight into the globalization impacts on our economy. The Statistics Canada I know is quite stuck in the mass production era of the economy and society from gender issues to productivity issues the dynamics that are fixed and fast frozen within those numbers in a way that does not reflect the changes that have been underway over the last 20 years. These are important constructs in determine new policies and making the necessary accommodations within our socio-economic landscape. To have one man do all this with a bit of an iron fist for that length of time is pure disaster. We live in a numbers based society with our scientific nose sniffing headlong into the twisting nether, and given that we need to have accurate dynamically accurate numbers and this we are a long way from. Whether it is the environment, wealth inequality, job quality, gender issues, the numbers are behind.

    They once in awhile pull off a high profile move such as an Aboriginal census and then use these way over due innovations and stake claims that they have become quite receptive to our needs.

    Ivan was the creator of the mother of all data cooking algorithms and that from my experience is the way he should be remembered as it is quite fitting really.

    Even the Center for health information decided to abandon Statcan’s organizational capacity and went their own way.

    They are masters of doing some high profile projects but at the expense of letting the core drift off into some quite debilitating statistical space.

  • In Straight Goods, Linda McQuaig writes, “StatsCan’s knack for dull presentation perhaps explains why the venerable institution has been able to survive and assemble much important information, even though its data sometimes embarrass its political masters. (For this, retiring Canadian chief statistician Ivan Fellegi deserves considerable credit.)”

  • Fellegi had a good parting shot in/at the National Post:

    Statistics Canada is ‘rigorously objective’
    National Post
    Tuesday, May 13, 2008
    Page: A15
    Section: Letters

    The National Post last week published a number of articles concerning Statistics Canada’s data on income and earnings from the 2006 Census and other sources.

    We are encouraged that one of our information releases has stimulated such vigorous public debate about important social and economic issues. We are also gratified that participants from all sides of the debate have made extensive use of our analysis. No one has questioned the accuracy or relevance of the agency’s information. We have worked hard to demonstrate the professionalism that justifies this confidence in Canadian statistics.

    What is disturbing is that Statistics Canada has been casually accused of bias and has been imputed with the most unsavoury of motives by a national newspaper such as yours — with no more justification than that some columnists disagreed with how we wrote our news release.

    I can only point out that, like most others, your columnists drew their own conclusions based on what we published. Indeed, Statistics Canada made available to the news media volumes of data, analytical publications and subject matter experts so that journalists could express whatever opinions they saw fit.

    Statistics Canada is rigorously objective in reporting the findings of its statistical programs. Reasoned criticism of our communication programs is expected and welcomed. But the recent comments published in the National Post were decidedly unreasonable and gratuitously damaging to our agency.

    Ivan Fellegi, Chief Statistician of Canada, Statistics Canada, Ottawa

  • I hate to be a stick in the mud, but I have a hard time with the progressive flag waving for Mr. Fellegi. He has the statistics act to stand behind, so I am not sure just how much he stretched beyond his political masters in providing a balanced numeracy type perspective of our socio-economic landscape.

    We have forever known the weekly car loadings of sulfur from just a few days ago, yet we still do not have an acceptable, statistically measured and supported means to evaluate poverty in this country. Am I just taking shots, no, far from it.

    I guess you had to work there to see the potential and the wasted opportunities to help inform rather than lather up and shine many of the prevailing pro-business and elite oriented manifestations within our little statistical world called life according to statcan. THere are other sign posts that stand in stark contrast to these, and the contradictions with their sharp edgesare but rounded and dulled off. Objective according to who, Ivan? I find it funny the right wingers are the ones who launch the assault on an already pervasively infiltrated organizational, “rule with an iron fist”, elite pro- business based mindset that dominates at the stats barn.

    Is it not so typical of the right, and apparently all those newspapers are just a bunch of socialist leaning reported too. I mean really when will the right stop their charade. I can only imagine the statistical agency without the statistics act and the reins given to the Harpers of the world. I would hate see the reports, so yes I do realize that some balance is sought but not nearly enough. And that brings me full circle, Ivan was in that position for way too long, and we should be insisting at this time that future appointments have a are much more periodicity focus. We need to make this point and to not make it at this time from such an forum as this in my view is highly irresponsible.

    Come on bloggers and PEFsters.

    We cannot take a chance on getting somebody with less balance, or we’ll never get rid of them.

    Mind you stability in statistics typically is a said objective. However I would much rather be inconsistently representative than being consistently non-representative.

    Given this age of reason and the numeracy that we seem to rely on to support many decisions effecting and defining the social economic landscape, makes this an imperative and important goal.

    I hope somebody takes me seriously and see my red flag!

    I guess you may have had to work there to internalize this, not sure.


Leave a Reply

Your email address will not be published. Required fields are marked *