Top

Honorary Member:

Barndorff-Nielsen, Ole E. (Denmark)

Elected Members:

Agboto, Vincent K. (Togo)
Al-Awadhi, Shafiqah (Kuwait)
Aldredge, J. Richard (USA)
Aoshima, Makoto (Japan)
Apaydin, Aysen (Turkey)
Argac, Dogan (Germany)
Bailer, A. John (USA)
Ben Arous, Gerard (France)
Bhattacharjee, Arnab (India)
Bolstad, William M. (New Zealand)
Bonzo, Daniel I.B. (Philippines)
Capilla, Carmen (Spain)
Carlson, Lynda T. (USA)
Chandhok, Promod K. (USA)
Chen, Jiahua (Canada)
Chikobvu, Perpetual (Zimbabwe)
Chinganya, Oliver J.M. (Zambia)
Cho, Wanhyun (Korea)
Cohen, Michael L. (USA)
Di Ciaccio, Agostino (Italy)
Doyle, Patricia (USA)
Eguchi, Shinto (Japan)
Ekol, George L. (Uganda)
ElSaadany, Susie (Canada)
Feng, Yuanhua (Germany)
Fielding, Antony (U.K.)
Fraire, Mary (Italy)
Han, Sunghee (Korea)
Hershberger, Scott (USA)
Hewitt, Linda (Trinidad and Tobago)
Higuchi, Tomoyuki (Japan)
Hooda, D.S. (India)
House, Carol (USA)
Hristea, Florentina (Romania)
Jeng, Shuen-Lin (Taiwan, China)
Jeon, Jongwoo (Korea)
Jyoung, Taik-Hwan (Korea)
Kamakura, Toshinari (Japan)
Karian, Zaven A. (USA)
Khattree, Ravindra (USA)
Kim, Ju sik (Korea)
Kim, Sulhee (Korea)
Kim, Yuntae (Korea)
Ko, Wang Kyung (Korea)
Kukush, Alexander G. (Ukraine)
Kurihara, Koji (Japan)
Kuriki, Satoshi (Japan)
Lamel, Joachim (Austria)
Lee, Jae J. (Korea)
Lee, Jae W. (Korea)
Lele, Suhbash (USA)

Top

An official presentation describing the research process and the report's main findings was made at the Statistics Netherlands Office in Voorburg, The Netherlands, to an audience of academics, press representatives, government officials, as well as Statistics Netherlands employees. The following interview, kindly provided by Mr. Eric Schulte Nordholt, project leader of the Dutch Census and one of the three editors of the book, provides some insight into this undertaking.

When did work first begin on the development of this virtual census?

A census is an event that occurs every ten years. The thought process behind the 2001 Virtual Census began during the Census of 1991. In The Netherlands, a traditional census had been common practise till 1971. Some problems experienced during the 1971 Census were:

Non-response: What can be done about non-response? For instance, about two percent of the population were non-respondents in the 1971 Census, should they be given a fine? It does not seem realistic to do so.

A debate regarding privacy came about, as the Government was busy introducing computers. It scared people into thinking "Big Brother is watching us"; therefore, a lobby group against the Census was started.

Cost: Is it useful to pay so much for a census? Besides cost, a traditional census is extremely time-consuming. By the time the checking, editing and analysing of all data has been completed, the figures are no longer accurate. The Monographs of the 1971 Census were published around 1980.

During the 1981 Census test, non-response was one out of four, thus it became clear that it was more difficult to interview people compared to ten years ago. In 1991, the Dutch Census law was abolished. Thereby, it would have been the end of the story for the Dutch Census that is if there weren't any international organisations, such as the European Union and the United Nations, who want to see census results of countries. In 1981 and 1991, results of the labour force surveys and information from the population register were only used for comparison with other countries, none of the results were published separately, and they were solely produced for the tables.

By the time the 1990’s came around, had the ideology switched back to either using the Traditional Census or a Mini-census? Both were rejected, as they were considered too costly. Dr. Cornelis van Bochove was the first Director who wanted to integrate the registers and surveys (micro-integration) at Statistics Netherlands by establishing the Social Statistical Database (SSD). Inconsistencies, such as when did unemployment start, were explored by the SSD. Through doing so, the SSD became stronger. It is very interesting what one can do with integrated data. In 1991, other countries published their census, but The Netherlands did not. For some countries extracting data is far easier than for others. For instance, the Scandinavian countries have more information in their registers and subsequently they can quickly produce their tables. At the end of 2002, I was asked to become the Project Leader of the Dutch Census at Statistics Netherlands. By that time, other countries had already published their first census results of 2001. I was able to set up the project team of ten colleagues by the beginning of 2003.

One important task for us was deriving variables. For some variables, we only had information from surveys. Two solutions were presented for this problem:

Weighting Technique: This technique has lots of restrictions and, thus, is not feasible.

Mass Imputation: The method by which you estimate as best you can the missing data. For a couple of variables, this is a possibility, but not for as many variables as we needed. With Mass Imputation inconsistencies can occur in the data set, for example married babies, which is visible in the tables.

In 1997, two members of the Statistical Methodology Department of Statistics Netherlands created a new method termed “Repeated Weighting”. This method was tested and, soon thereafter in 2000, the software was successfully built.

How can one compare data from a sample-based census to that of a pure census?

For a sample-based census, the sample can be large enough to collect all your data, which means less overhead cost. However, you cannot use it to correct your registers. As for a traditional census, you obtain all your information and consequently tabulate all the data you want; however, some variables may still remain unknown. One important breakthrough was due to a change in the Dutch law. Statistics Netherlands was thereby given the ability to use the registers that already exist in the country. Not all countries' statistical offices have this luxury. The majority of the European nations, even the United States and the United Kingdom, conduct traditional censuses.

What are the potential weaknesses of a virtual census?

There are only a handful of countries that do not conduct the traditional interview. The Netherlands is unique with its virtual census, whereby the non-response problem has virtually been eliminated. On another note, the privacy issues have disappeared as the society's attitude has changed since 1971.

Do the register-based data (Social Statistical Database) correspond to the previous census data? Were adjustments necessary? Which additional sample surveys were necessary (e.g. education level, occupation)?

You can compare data with other countries, as well as over a period of time. No special adjustments were made. The 2001 Virtual Census is more of a headcount in order to see how strong we are as a nation in certain areas. For instance, economic status was an area we explored in terms of whether citizens are e.g. employees or self-employed. During interviews, mistakes can be made, such as the selectiveness of the response. Some variables are not even answered so you would have to correct the data. Still some information was needed from surveys, because not all variables are available from the registers, such as hours worked per week, educational level, occupation; thus, information from surveys was used.

Can you explain the estimation method ‘Repeated Weighting’?

Repeated Weighting is a type of weighting technique. The weight can change depending on the table you need to fill. The European Union requires forty tables in total: eight tables on living, two commuting tables, while the remaining thirty tables are interlinked and produced using Repeated Weighting software. So, it only took one year to complete the 2001 Virtual Census due to the fact that the software had already been set up. The software had been used in two earlier surveys in 2002. My team are an experienced group of people, including a population specialist, employment statistics (labour force) specialists, who experienced a lot of learning while on the job.

How do you adjust for non-response between traditional census methods and this new method?

During a traditional census, non-response has never been a problem in the past because people were jailed and/or fined for non-participation. Later on, non-response became a major problem. With a virtual census, non-response is adjusted with the Repeated Weighting technique.

What was the cost of this Virtual Census in comparison with a traditional census?

A traditional census can easily cost a few hundred million Euros. The 2001 Virtual Census cost only a few million Euros, about 1% of a traditional census. People say that it is unfair to compare because during our Virtual Census we already had use of the Dutch registers, and the cost of obtaining the registers’ information is not included. If you had to create registers for a virtual census, then a virtual census would become more expensive than a traditional one. I would not recommend doing a virtual census if the registers did not already exist.

What are the most significant results of the Census in terms of comparable results?

That depends on your interest. Generally, remarkable outcomes are structure, relation between variables, and disaggregation. More people of Dutch origin work than those of Surinamese origin in The Netherlands, but more Surinamese women work than Dutch women. There are enormous opportunities presented with these data. Once you have the information, you can correct for it. With the results of the data, you cannot derive one-liners as is done ever so often by the press, you need the story behind it. I'm not giving a solution; I'm just simply stating the facts.

What steps were taken to ensure consistency in the data input?

Get the input from different sources. We used the micro-integrated data sets from Bart Bakker's SSD group, so that the consistency of these data was guaranteed. Some surveys, which had already been conducted by Statistics Netherlands, were used, but in order to keep the consistency, Repeated Weighting was used.

What are the margins of accuracy (in the virtual census)?

We want the accuracy to become 100%, but that is not realistic. A main group is missed: the illegals, we know they are here, but they are not in the registers. On the other hand, they also would not participate in a traditional census. There are about 100,000 illegals in The Netherlands with large margins. That is a small percentage, less than 1%, of the population, which makes a difference but not a major one. Historically, there are two types of population. Dejure (used in all our 20th Century Censuses) are the people in your registers, hence the type of population used in our Virtual Census. Defacto (used in some of our 19th Century Censuses) are the people present in the country at the time of your census, so those on sabbatical, for example, are not included. The illegals are a major problem. But for distribution, if you miss less than 1% of the population, then this group will not be a big problem. One question that still remains is how accurate is the information? Information contained in the registers is counted, so therefore completely reliable. With the software for Repeated Weighting, you can also calculate the margin of error. If there is too little information per cell, such as a weak estimate, it is risky and therefore it is not published. That is a main problem of survey estimates.

How comparable is the Dutch Virtual Census of 2001 to that of other EU countries? Have the EU countries used similar variables?

The EU subsidiarity principle is every decision that can be made at a national level, should be made at a national level, as long as a country can fill most of the forty EU tables. Even in a traditional census, some cells may be missing. Also, how much effort does a country take to answer a specific question? Although my team started late on this project, we were one of the first countries to produce the complete set of tables. All countries had to produce the forty tables with the same variables. Countries do compare their results with each other. Eurostat is making a complete final comparison, but they have to wait for all countries to submit their statistics. However, they must be careful with certain comparisons; for instance, there are different educational systems in each country, which in spite of the harmonised classification, may not be quite comparable.

If you were to undertake this project again, based upon your experiences, what would you do differently?

The EU census occurs every ten years. By the time the 2011 Census comes around, the baby-boomers will have retired; there will probably be a different Director General at the statistical office, so many changes will have occurred between now and then. New, young people will have to be hired by that time, but they will need to gain experience first. Many statisticians do not yet know about Repeated Weighting. This Virtual Census was conducted in preparation for the future. For the next time, there are some foreseeable risks, such as decreased members of staff due to budget cuts. Another risk due to budget cuts is the detail of the surveys. Will there be fewer details available? Organisational changes will also have to be factored in; for example, the Statistics Netherlands’ Voorburg Office will be moving in 2006, but where to? This all has an effect on the next census. For the future censuses, we need to try to add some variables to our registers, such as the level of education and occupation. So there are challenges to prepare ourselves for the next round! 

Top

Celebrating the 350^th anniversary of Jakob (or Jacob, or Jacques) Bernoulli’s birth either late in 2004 or early 2005 depends on your taste in calendar reform. Bernoulli was born in Basel into a wealthy Protestant family that two generations earlier had fled Amsterdam to escape enforced conformity to Roman Catholicism. The Protestant cantons of Switzerland did not recognize Pope Gregory XIII’s calendar reform of 1582 until 1701. Hence, Bernoulli was born on December 27, 1654, under the Julian calendar as recognized by his canton, or on January 6, 1655, under the Gregorian calendar.

Forced by his parents to study philosophy and theology, he obtained a Master of Arts degree in 1671 and a Licentiate in Theology in 1676, both from the University of Basel. At the same time, and against the wishes of his parents, Bernoulli also studied mathematics and astronomy. His was the first of at least three generations of the Bernoulli family to work in mathematics. His brother, Johann Bernoulli (1667-1748), also became a mathematician. Although the younger Bernoulli was taught his mathematics by Jakob, the two quarreled in the 1690’s and became rivals. One of Jakob Bernoulli’s biographers has described him as, “self-willed, obstinate, aggressive, vindictive, beset by feelings of inferiority, and yet firmly convinced of his own abilities”.

After graduation in 1676 from the University of Basel, Jakob Bernoulli worked in Geneva as a tutor. He subsequently traveled to France, the Netherlands, England and Germany, returning to Switzerland in 1683 at which time he took up a position at the University of Basel teaching the mechanics of solids and liquids. He had been offered an appointment in the Church, reflecting his university training, but he turned it down to pursue mathematics. In 1687, Bernoulli was made Professor of Mathematics at Basel and remained in that position until his death in 1705.

In the wider arena of mathematics, Bernoulli made important contributions to algebra, infinitesimal calculus, the calculus of variations and the theory of series. Much of his work was inspired by his researches into optics and mechanics. It is, however, in probability theory that he is best known. His fame rests on his most original work, the Ars Conjectandi, edited and published posthumously by his nephew Nicholas Bernoulli (1687-1759) in 1713. This highly influential work took probability in new directions, taking the initial concepts and results, which had been established by Pascal, Fermat and Huygens in the 1650’s, and proceeding to interpret probability as a measure related to certainly, necessity and chance. Applications of the theory went beyond the enumeration of outcomes in games of chance to such new areas as the interpretation of evidence. There were also extensions to the enumeration of chances by the inclusion of new combinatorial results, such as the development of Bernoulli numbers. Bernoulli formally introduced the idea of estimating probabilities through relative frequencies of observed outcomes and obtained an associated theorem, now known as the Bernoulli’s weak law of large numbers.

On his death in 1705, his brother and rival became Professor of Mathematics at the University of Basel.

David Bellhouse
Chairman
Christiaan Huygens Committee on the History of Statistics 

Top

You were born in Hungary and obtained your B.Sc. from the University of Budapest in 1956, the same year as the Hungarian Revolution. Could just tell me something about that time and your coming to Canada?

I had the misfortune to be born into a well-to-do bourgeois family. Back in the early 1950’s in Hungary, people from my class were not permitted to attend high school, never mind university. I managed to escape this fate in 1949 when my family moved from the small town, where I was born, to Budapest. Things were not well organized there, and so I was able to get away with the larceny of enrolling myself in high school. University would have been totally out of the question with my background if it hadn’t been for a loophole, perhaps placed there precisely for undesirables like me. There was a national competition in every subject taught in high school, and those who scored in the top five in the country in each subject were guaranteed a place in university, irrespective of their background. And so that was my only possible option. I could choose only one subject to compete in, but which subject? Since I was planning at the time to become a poet, my teachers encouraged me to enroll in Hungarian literature as my subject. After I mulled it over, I decided that literature was not really a smart choice because they could always say they didn’t like what I wrote, and how could I argue against it? There was only one subject where there was no argument: mathematics. So I enrolled in the mathematics competition and managed to come in third in the country. With that and a lot of good connections, I got into the mathematics faculty in Budapest. That is how I became a mathematician. 

What about coming to Canada?

In 1956, during the Hungarian revolution, I seized the first opportunity to get out of Hungary. The Canadian government was very generous, admitting 30,000 Hungarian refugees. I became one of them, and joined my sister in Ottawa. She and her husband were studying medicine, and they suggested I do the same. I thought that this sounded reasonable, so I registered at the University of Ottawa.

You said you escaped. Did you have to walk? I know of other people who have walked out and into another country.

First, my cousin and I took a train into the mountainous border region of Hungary. My mother had managed to get me a medical certificate that said I was suffering from tuberculosis. I didn’t have TB, but a sympathetic doctor—who knew exactly what this was all about—gave me a referral to a sanatorium conveniently located in a border town. When we reached this town it was late afternoon. Because of the curfew there, it was too late to go up to the sanatorium, so we got permission to spend the night in the local hotel. This was convenient because the next morning we were supposed to meet somebody who would guide us over the border. When no one showed up, we just set out on our own and walked, carrying our belongings in rucksacks. When we encountered an old woodcutter along the way in the mountains, he looked at us and said, “I don’t know where you are going nor do I want to know. But that way [pointing in one direction] there are Russians. This way [pointing in another direction] there are no Russians”. So we walked ‘this way’ and managed to get to the border. I didn’t believe it was the border until I heard an Austrian border patrol. They picked us up and took us to a camp, and from there we went on to Vienna.

And then to Canada. What an amazing story!

I arrived in Ottawa in early February 1957. Ottawa University admitted me to medicine, but I had several months to wait and I wanted to earn some money. Some of my sister’s friends worked at Statistics Canada, then called the Dominion Bureau of Statistics. They knew I had a mathematics background, and they encouraged me to apply since there was a shortage of mathematicians at the time. I applied and was interviewed by Doug Dale, a section chief who subsequently became a professor at Carleton. I could hardly understand what he was asking me, but I must have nodded and shaken my head at the right times. More important, though, were my academic transcripts, which were signed by some very famous names—my professors in Budapest. I think those transcripts probably contributed most to my being hired by Statistics Canada for the princely sum of $3,300 a year as a technical officer, which was a glorified clerk. 

So your first job was in Statistics Canada or, I should say the Dominion Bureau of Statistics?

It was temporary until the fall because I planned to enroll in medicine. But by mid-summer, I was having second thoughts about medicine because I really liked what I was doing with statistics. Finally, I went to Carleton University and said, “Looking at your calendar and my transcripts, I think I have actually met all your course requirements for a B.Sc., so I want to enroll for my master’s”. This involved quite a bit of a ‘chutzpah’ on my part, since I hadn’t quite finished my first degree in Hungary before I escaped! But Carleton, a small and non-bureaucratic university at the time, said, “Well, that sounds reasonable. Are you willing to stand for an examination for your B.Sc. material?” I said, “Sure”. What did I have to lose? Three or four professors examined me, and shortly after that I was admitted to a master’s program.

So while working at Statistics Canada, you studied for a master’s at the same time?

Yes, they allowed me to make up any work time I missed. At Carleton, I was the only master’s student at the time. In fact, I was the first master’s student in any subject in their new graduate program. I went mostly in the evenings. There was no formal instruction as such: we mostly discussed reading materials. It was much like the British system, where a reader helps you understand the assigned readings. There were also some projects to hand in. It was really a very personalized situation. I proceeded at my own pace, and it couldn’t have been more luxurious.

And during the day, what were you doing at Statistics Canada? You started off, as you said, a glorified clerk. 

I was still doing glorified clerk work, but I found some patterns that interested me. Actually, I still remember the project. I was comparing the size distribution of city blocks in various cities. There were big folders of data, block by block, to be compared with the size distribution of the selected city blocks. They were really very far apart and there was a bias in the selection, which is what I detected by looking at this pattern. That provided my master’s thesis topic, which I ‘mathematicated’ sufficiently to make it a mathematics thesis. Carleton didn’t have a statistics professor at the time, so my master’s degree had to be in mathematics.

Then you went on to the Ph.D. program. Was there then any statistician at Carleton?

No, there still wasn’t, and the course content was mostly mathematics. But I wanted to write my thesis in statistics and they made an arrangement with Don Fraser at the University of Toronto to be the external examiner. As I was also Carleton’s first Ph.D. candidate, they didn’t want to be accused of giving out a cheap degree, so they brought in three professors from Ottawa University and three from Carleton to give me an oral exam.

Were there any statisticians at this oral examination?

No. It was an oral in mathematics. The statistical component was the thesis, and that was handled by Don Fraser.

And the thesis was Sampling with Varying Probabilities Without Replacement: Rotating and Non-rotating Samples, which I assume is a Statistics Canada problem?

It was a Statistics Canada problem. By that time, I was in charge of redesigning the Labour Force Survey. One of the issues was how to select with probabilities proportional to size without replacement, particularly where there is a rotating sample and you want to retain as many from the previously selected sample as possible because it is expensive to throw them away. That became my thesis project.

Is it related to your first paper that was published in JASA?

Yes. I wasn’t completely happy with the thesis and so I spent more time and did some work. Then I submitted it to JASA. 

So how did Statistics Canada fit in this whole thing at the time? They allowed you to take a Ph.D. Were they supportive?

At that time, Statistics Canada didn’t have a very intellectual environment - quite different from what it is now - but it was supportive. Here was a Hungarian refugee who looked like he needed help, and they were really very nice about allowing me to make up the time I spent on my studies. In those days, a bell rang at the start and end of the working day. If you were absent three times when the bell rang, there was a deduction from your salary. It was a very rigid environment, and not just here, but everywhere. So letting me have time off and be on the honour system to make it up - after hours or at home, when no one could supervise it - was very unusual. 

In 1962, you were promoted to Chief of the Sampling Research and Consultation Unit. You became Director of Sampling and Survey Research in 1965; then Director General in 1971 for Methodology and Systems; then Assistant Chief Statistician in 1973 for the Statistical Services Field; and along the way you kept writing papers. As you progressed through different areas of Statistics Canada and took on new work that inspired new statistical research for you. 

Yes, that’s certainly the case. I benefited from three things in my early career. Very early on, in 1958, I came to Nathan Keyfitz’s attention and he took me on almost as a protégé. He was certainly the greatest statistician in the Dominion Bureau of Statistics at that time. Before he left Statistics Canada, he made a very conscious effort to organize a series of meetings with every director, at which he asked me to be present. The two of us conducted a review of potential sampling applications in Statistics Canada. That was really an enormous career boost, not just from my association with the great Nathan Keyfitz but also in terms of the insight that it gave me. It was a unique opportunity to review every potential sampling application in the Bureau. In those days, the only actual sampling application was the Labour Force Survey, so this was really Nathan Keyfitz’s parting gift to Statistics Canada and to me—to encourage the use of sampling much more broadly. When he left, Simon Goldberg, an assistant chief statistician who was one of the true founding geniuses of Statistics Canada, continued to guide me and supported my career. Most importantly, he introduced me to Morris Hansen and Bill Hurwitz when they came to Ottawa. My real postdoctoral training came when Hansen invited me to be a member of the Methodology Advisory Committee in the U.S. Bureau of Census. On that Committee, which was chaired by Bill Cochran, were Nathan Keyfitz, Bill Madow, H.O. Hartley, Fred Stephan and little Ivan Fellegi. In addition to those Committee members, there was a galaxy of absolutely top-notch people in the field: Bill Hurwitz, Max Bershad, Joe Daly, Margaret Gurney, Joe Waksberg and, of course, Morris Hansen. That was the golden period of the Census Bureau and there was no better place, really. That was how I learned statistics. I read a lot but never attended courses, because there was nobody who could teach me at Carleton. In fact, the first statistics courses I took were ones that I was giving.

So now you’re in charge and you want to be able to encourage this kind of activity in Statistics Canada. How do you go about it?

First of all, I believe the most important thing I have contributed to Statistics Canada is an environment that encourages intellectual curiosity. That is part of my own makeup for sure, and why I ended up writing papers. They were always motivated by something that needed to be solved and, somehow, thinking about the underlying theoretical problem made the empirical problem a lot more orderly. Problems manifest themselves in very disorderly fashion but very often, if you can find their underlying logic, they sort themselves out. So it was that desire for understanding that prompted me to do research myself and encourage anybody who worked for or with me. Secondly, there are more administrative sorts of tools. We provide what we call ‘block funds’ for research. Most of our budget is allocated to projects that have a certain output that is supposed to be achieved by the end. Over and above that, however, we provide these block funds to certain professional groups for research. These funds are not tied to any particular project and are designed to support research on topics the researchers and their peers think are reasonable. Another incentive for research is attending conferences: your paper is your ticket. There’s no guarantee it will be accepted, but if you don’t write a paper you are almost guaranteed not to go. So it comes back again to the environmental question. There is no magic. You do it by personal example if you can, by encouraging others indirectly, and by creating the economic means and some incentives in addition. If you hire good people, it takes off. 

As your career has progressed, there has been an evolution in the papers you have written. You started off with mathematical papers and now are writing more philosophical papers about the nature of organizations. Do you miss any of the mathematics?

I do, but there are offsetting benefits. I still “mess around” with data, but in a different way. I believe that it’s very much a function of the Chief Statistician to make presentations about what society is about in Canada, based on our own data. I’m no longer looking for methodological irregularities, but seeking explanations of the phenomena around us. Many more people can write methodological papers than really understand how to run a statistical office, so I can make as Chief Statistician a greater contribution at this point. But I do have nostalgia for the ‘good old days’ when I had the wonderful feeling that I could do the job of anybody who was working for me. I don’t have that feeling anymore. 

How would you characterize the theme of the papers you’re writing now?

They are about what makes good statistical offices function well and what some of the challenges facing them are.

Some of the key words in those papers are things like credibility, trust and confidence. What do you see as the major risks in this area in Canada or generally in statistical organizations worldwide?

I think trust is a fundamental issue for statistical offices. I believe strongly that there are very few people who can directly test whether official statistics are reliable and credible. And that is a fundamental vulnerability because without proof, it really boils down to trust in the provider. You can’t ‘kick the tires’ directly. The information may be good, but it won’t be used if it’s not trusted. So trust really is fundamental: how to gain it, how to nurture it, and how to make sure we don’t lose it. 

A number of things create trust. Obviously, good methodology is the foundation. Beyond that, there has to be a true arm’s length relationship with government. The lack of this relationship is every bit as pernicious as bad data would be. But the existence of an arm’s length relationship isn’t always obvious, so we have to make it evident by what we publish and the kinds of analysis that we do. One of the reasons I put so much emphasis on doing analysis of our own data to highlight findings is because analysis is what the public can consume. Another reason is that analysis is what gets media attention. The kinds of analysis that you put out can signal that you are genuinely independent. So does controlling your own budget without any government interference in how you spend it. If you don’t have trust, as a statistical agency you might as well go out of business. If you gain trust, then you are given tasks - and the money to go with them - that you wouldn’t get without that trust. 

Let me give you two examples of what I mean about trust. A few years ago, the Federal Government wanted to harmonize the federal GST with provincial sales taxes so that businesses wouldn’t have to charge two different amounts. They managed to get an agreement with three of the Atlantic provinces to harmonize, but the provinces insisted on first knowing the amount they would have received through their provincial sales taxes - something that could only be done through a simulation involving an enormous and complex economic analysis by Statistics Canada. The Provincial Governments trusted us to do this analysis, even though it involved billions of dollars and the results couldn’t be verified. Because of their trust, we got a lot of extra money to expand the detail, the scope, and the reliability of our economic statistics program. 

Another example is when the federal government transferred to the provinces, on two different occasions, tens of billions of dollars for health care. While the provinces could spend the money as they wished, without federal interference, the spending had to achieve certain objectives that would transform the health system. Statistics Canada was approached to develop many of the health statistics that would show whether or not such a transformation had occurred at the end. Again, we got a substantial amount of money that has ultimately expanded the ability of the public to monitor the performance of our health system. That was a question of trust in a domain that was entirely politicized and still is, but in which everybody accepted Statistics Canada as an honest agent. There is an increasing desire in society for honest brokers to play a societal role of arbitration. In our case, of course, it’s arbitration in terms of providing credible information that is trusted. I think that in this respect, we play a hugely important social role that no other statistical agency in the world, to the best of my knowledge, has managed to play to the extent that we do. And it’s uniquely as a result of the trust and of our ability to not just gain the trust but also make people aware that we are trustworthy.

Let’s go back a bit earlier in your career. Back up for a second. At one point along the way you left Statistics Canada in the sense that you were seconded to Washington D.C. to sit on President Carter’s Commission on the Reorganization of the U.S. Statistical System. What was the attraction? Why did you take it on?

There was a push and a pull. In the late 1970s, Statistics Canada was being torn apart by deep management problems and issues. I don’t want to go into details, but I was in fundamental disagreement with the policies that the management wanted to pursue at that time. I took a leave of absence, but my mind was more or less made up that I would not come back. The opportunity to act as a kind of staff director of this presidential project in Washington presented itself just at the right time. The totally decentralized U.S. statistical system is quite different from the centralized system we have here in Canada. I thought it would be a wonderful opportunity to look at the polar opposite of the Canadian system.

So what different perspective did it give you? 

It crystallized for me the advantages and disadvantages of centralization versus decentralization. A decentralized statistical system is the ‘minimax’ solution, but it’s certainly not the optimum solution with respect to efficiency or effectiveness. It’s minimax in the sense that if the one statistical agency the country has is really bad, everything is bad. That can’t happen in the United States because they minimize that maximum risk through decentralization. I’d say that’s the great advantage of decentralization. 

I learned something else in the United States. I went down with the bias of a methodologist, believing that the world thought the U.S. Census Bureau was a better statistical agency than the Bureau of Economic Analysis and the Bureau of Labor Statistics. But in Washington, I found that the users didn’t know much about the Census Bureau except it did the census. They were largely unaware that it also carried out a very large proportion of the current U.S. surveys. After several months, I came to a very clear understanding of why this was so: it was because the Census Bureau did very little analysis. It produced wonderful statistics, but nobody used these directly except the cognoscenti. The public obtained most of its information via the analyses produced by the Bureau of Labor Statistics or the Bureau of Economic Analysis. As far as the Washington elite were concerned, these agencies had the people thinking that the Census Bureau were the “plumbers”. That’s when it crystallized in my mind that analysis is crucial. You can do the world’s best data collection, but it’s the analysis that gains you prestige. 

And that’s something you brought back with you to Statistics Canada.

Absolutely. I was asked to come back to Statistics Canada in 1979 because the 1981 Census was in trouble. I came back in a different role from when I left: I made the switch to subject, as opposed to methods, and took on social statistics, which included the census. My motivation to come back was to save the 1981 Census if I could. But I also had the opportunity to put into practice what I had learned in Washington in the social domain: enhancing our analytic output to make it more readable. Our analytic output has increased by orders of magnitude over the last 20 to 25 years, and I could claim some personal credit for that. 

We’re getting near the end here, so I have a couple of general questions. One is, again from reading the Canadian Who’s Who, I notice that you are 68, three years beyond normal retirement age. What keeps you going in the job and what do you like about it that keeps you going?

Well, I can almost turn it around. Why shouldn’t I?

I’m asking the questions … 

I think I’m really one of the world’s very fortunate people, because I have a job that has a unique window onto practically every aspect of Canadian society and the economy and life in general. There is hardly an aspect of Canadian life in which we are not somehow involved - labour markets, production, sales, immigration, education, health (not just health in general but the determinants of and the outcomes from interventions) and education. One of the unique parts of being Chief Statistician of Canada is being able to set priorities. Through the setting of priorities, I can actually have an impact on what we illuminate, where we put emphasis. That is singularly interesting and satisfying, and it puts me into contact with a lot of wonderful people. You can’t set priorities unless you gain a certain level of understanding of the various places where you can put your money and what benefits might come about as a result of putting some additional resources here, there or elsewhere. It’s a fascinating job. 

The Canadian Who’s Who says you’re also interested in classical music and reading. Who are your favorite authors and composers?

I have a long list of those. I like to read at least two books at the same time. One is usually history, and the other is usually fiction. Over the last year’s reading—I can’t go back beyond the last year because it’s too long a list—I really loved Jared’s Guns, Germs and Steel. I also read Margaret MacMillan’s Paris 1919, a wonderful, well-written book. I enjoyed a biography of Khrushchev by Taubman and a book by Norman Davis on Europe. When it comes to literature it’s a very long list of authors. I love Tolstoy, Gogol, Thomas Mann, Flaubert, Stendahl …

What about music?

Music? Again, it’s a very long list, but it certainly includes Mozart, Beethoven, Bach, Brahms and Schubert. Of the more modern musicians, I love the Russians - Shostakovich and Prokofiev. And I really love Mahler and Bruckner. In fact, if I had to choose a favourite, it might be Mahler.

Well thank you, it’s been fascinating.

You are most welcome. 

Top

Marco Polo Committee on Statistics of Travel and Tourism 

The idea for a committee devoted exclusively to the discussion and promotion of tourism statistics first emerged in 1991. That year saw two events that stimulated awareness of the need for such a scientific forum that was to become the Committee on the Statistics of Travel and Tourism (COSTT). The first of these was the organizing of an Invited Paper Meeting on travel and tourism statistics, which was organized at the 48^th ISI Session in Cairo. The Meeting was well-received and generated discussion on the need for continuing networking among tourism statistiticians. Second, the World Tourism Organization organized the “International Conference on Travel and Tourism Statistics”, in Ottawa, Canada. A number of key definitions and concepts related to tourism statistics were achieved at that Conference, as well as the call for the implementation of improved statistics and analytical techniques.

The 49^th ISI Session in Florence formally recommended the creation of the Committee on Statistics of Travel and Tourism, with Raphael Bar-On (Israel) as its first Chair and Enzo Paci (WTO), Vice-Chair. An Executive Sub-Committee was also struck to guide the administrative requirements of COSTT. 

The general aim of COSTT is to improve the collection, analysis, publication, and use of statistics on tourism demand, supply, and other economic measures. The Committee brings together both producers and users of tourism statistics.

Members come from diverse backgrounds, but are primarily associated with national statistical organizations, national tourism bodies, international tourism or statistical organizations, and universities. All share a technical or scientific interest in improving the quality of travel and tourism statistics, and in promoting greater consistency and comparability among international tourism statistical sources.

In 1994, the Committee was also given the name Marco Polo Committee, after the legendary early traveller. The Marco Polo Committee held its first Invited Paper Meeting in conjunction with SCORUS (International Association for Regional and Urban Statistics) in 1994. This was followed by COSTT organizing its first ISI Invited Paper Meeting at the 50th ISI Session in Beijing. 

Invited Paper Meetings have been held at all subsequent ISI Sessions, and the Committee continued to attract tourism statisticians from Europe, Asia, North and South America, Australia and New Zealand. The growth of COSTT resulted in the negotiation of a partnership with the Canadian Tourism Commission as its donor agency to fund a COSTT Secretariat at the Centre for International Studies on the Tourist Economy (CISET), the University of Venice. 

At the 53^rd ISI Session (2001) in Seoul, COSTT arranged an Invited Paper Meeting on household surveys of domestic and international tourism travel patterns.

Most recently, at the 54^th ISI Session (2003) in Berlin, COSTT organized two Paper Meetings: an Invited Paper Meeting on tourism and transportation satellite accounts, and a Contributed Paper Meeting on urban and regional tourism statistics. 

In 2001, COSTT Chair Scott Meis organized and hosted the first international conference on Tourism Satellite Accounts in Vancouver, Canada. A number of Marco Polo Committee members spoke, both as Invited Speakers and as authors of Contributed Papers. Several COSTT members were also recruited to provide papers to the first issue of the “Enzo Paci Papers” on methodological topics in tourism statistics. The collection was published by the World Tourism Organization and, in 2004, will publish the fourth volume. COSTT members continue to play a central role in providing papers for this annual collection.

In its initial six years of existence, with limited resources, and a specialized mandate, COSTT has positioned itself to work collaboratively with larger established scientific organizations devoted to tourism research such as the World Tourism Organization, the Association Internationale d’Experts Scientifique du Tourism, Travel and Tourism Research Association International, Organization of Economic Co- operation and Development, the Tourism Working Group of the Asia-Pacific Economic Co-operation, Pacific-Asia Travel Association, the Society for Opinion and Marketing Research and the Tourism Committee of EUROSTAT.

An enduring issue that motivates much of the activities in COSTT is the recognition of the need for further improvements in the quality of international tourism statistics. Industry advocates claim that tourism is the world’s largest industry. Until recently, though, the development of travel and tourism statistics and analytical tools lagged far behind the emerging economic and political significance of the phenomenon. An international consensus on how to even define tourism (as well as the various sub-classifications of domestic and international tourism) did not exist until the 1991 WTO Conference mentioned previously. 

Over the last decade, a revolutionary breakthrough in the analysis and reporting of tourism statistics has been achieved – the Tourism Satellite Account (TSA). This tool permits the measurement and tracking of tourism in ways that are consistent with those used in conventional industries by National Accountants. The emergence of the TSA has created an explosion in the demand for improvement in the availability, quality, and coverage of tourism statistics. COSTT members are at the forefront of working to help nations and statistical organizations meet this demand.

The work of COSTT will once again be presented at the 55^th ISI Session in Sydney. An Invited Paper Meeting on international travel surveys has been organized and will feature speakers from a number of countries who will describe their nations’ experiences in designing and implementing surveys to collect data on international travel patterns, especially on expenditures by international visitors.

Stephen Smith
Chairman
Marco Polo Committee on Statistics of Travel & Tourism