Close
UCL Minds 0:07
You're listening to a new series of the UCL Statistical Science Podcast. In this conversations we will speak with accomplished and interesting UCL alumni who have worked across a wide spectrum of applications and industries, and we share their career stories, achievements and advice. Whether you're an aspiring statistician or simply curious about the world of data science, this podcast is for you.
Nathan Green 0:41
Hi, everyone. I'm Nathan Green. I'm a member of the department. I'm very pleased to be joined today by Michael Baxter, who's an alumni of the department. And he's gonna be telling us about his time at UCL and what he's been doing since which is a lengthy list. I'm very much looking forward to that. So before we get into all that, would you mind briefly introducing yourself Michael?
Michael Baxter 1:06
My name is Michael Baxter. I took my MSc in UCL, I graduated in 1977. I was at the time the youngest person ever to get an MSc with distinction in statistics, because I arrived at UCL 17 and got my BSc when I was 20. And I got my MSc on a civil service bursary, and I was in the civil service for many years. And I've never left the civil service and I'm working for a market research firm called Cantor.
Nathan Green 1:37
That's a tease for some of the things that we can talk about. So you said you're at UCL in what year?
Michael Baxter 1:44
Well, I took my BSc from 73 to 76. That was in the maths department. Yes. And I did my Masters 76 to 77.
Nathan Green 1:52
And what was UCL like at that time?
Michael Baxter 1:54
It was a very good place to be. It had very high standards, students are well disciplined. We didn't have any of the rowdy ism that's universities had at the time and got a sense of history. The math department was actually the oldest building that occupied was predated the main building. In the maths department we had to it was very distinguished lineup, we had two Fellows of the Royal Society is the joint heads, plus another two, who were emeritus professors still knocking around. And I've taken a math degree because that was my good subjects. I haven't really given much thought to what I'd do with it. My father was a civil servant, he said, after become a civil servant, but not as a mathematician, but as a statistician because they were much better treated. So I applied to be a government statistician. And they said, Well, yes, you're a very suitable person, but you don't know enough statistics. Would you like a bursary to do a Masters in statistics and fine, I thought, well, I know UCL and I know it's got a very good statistics department. I'll go back there. So I did that. That was a bit of a mistake, because the MSc tended to assume you already had a degree in statistics, whereas most statistics MSc is assumed degree in Maths. But I enjoyed the challenge.
Nathan Green 3:10
A challenge you met by the sounds of things. So what sort of things did you do in the MSc?
Michael Baxter 3:14
I did a range of subjects, so to time series, and Bayesian decision analysis, index number theory, which was very useful where we took on the retail price index later, I did my thesis on the properties of the Pareto distribution as supervised by Rex Galbraith and I extracted some of the results of that and I had it published in Metrica. Its had I think, 56 citations according to Google Scholar, that was actually the first refereed paper I had published. The second one was about Lewis Carroll, not about the Pareto distribution, no, had a lifelong interest in Lewis Carroll. And it's well known that many of his poems are parodies of poems that have been well known to his audience. And one of his poems that was assumed to be just sort of general burlesque of the style is the border ballad. I was able to demonstrate that it was a close parody of one border ballad, in particular, and I even demonstrated that there were several published versions in which published version was the closest to what he did parodies.
Nathan Green 3:16
That's interesting. So where did you get that published then?
Michael Baxter 3:47
The Journal of the Lewis Carroll Society.
Nathan Green 4:25
Right. Okay, that makes sense! Very good. So was it usual for Masters students to publish papers?
Michael Baxter 4:33
No, but Rex Galbraith thought it was well worth publishing.
Nathan Green 4:37
Yeah, it sounds like it. What were people's perspectives on Bayesian statistics at that time, you mentioned Bayesian stats.
Michael Baxter 4:43
The department was very devided to the head of department, Dennis Lindley, he was of course, the leading Bayesian Statistician in Britain at the time, that the government Professor, Mervine Stone, was much more sceptical about Bayesian ism. And we used to have the journal Club, where all the lecturers and the post grads got together and discussed things. And we had some interesting debates about Bayesian statistics.
Nathan Green 4:43
Literally you said you had two heads of department and one of them was Bayesian and the other not so you were literally split. Oh, interesting. So I've got no idea what UCL was like in the 70s. The department was in a different building right, on the left as you come into the main quadrangle. So nowadays, we're not in the main quad. We're out on the side streets. It was interesting, you said that your dad said you can't be a mathematician you have to be a statistician in the civil service. And I kind of always thought that statistics was sort of looked down upon relative to mathematics til relatively recently.
Michael Baxter 5:08
The way the civil service system is structured, is mathematicians were scientific officers, and statisticians were a separate class, statistician, and we were fast streamed. The idea was that we will get to what's now called grade seven, in a much shorter time than most civil servants with most civil servants never get to grade seven.
Nathan Green 6:07
Yeah, so it sounds like you excelled in your time with the BSc and your Masters, then what did you start working on in those early days?
Michael Baxter 6:16
Population statistics, first, I was doing projections, we were skipping much more than forecast of the population of individual counties and major cities are important for town planning things. Then I was moved on to population estimates. Nowcasting, if you like. And in 1981, I was sent to the Bureau of the Census in Washington to find out how they did population estimates. And they said along the whole, they didn't have much to teach us because it had to develop all sorts of clever statistical techniques, but it's because they had lousy data. We had much better data, we said we wouldn't need these techniques. But one thing I learned from them, they just taken the 1980 census, and they showed me how they had checked the census and particularly how they estimated the undercount. So when I got back home, we take the 1981 census. And I thought, would be interesting to try and apply these American techniques to see the undercount of the 81 census. And that was important because we'd have to revise all the population estimates back to 1971. to emerge right at the senses, and I told my boss, the undercounting in the in our census 81 and also 71 is much bigger than the census people are claiming. And my boss said write this ups and tell them. So I did that and they weren't very amused, he was wonderful boss, Terry Orchard. And he persuaded everybody that I was right. I can't remember the exact figures, I think I think they should be under counted something like 40,000. Actually, it was 250,000, which had a population of 50 million is not enormous enough to affect the biggest reshape. So that was the official population centres and surveys, which is now part of the official national statistics. And then in 1983, I moved to road accidents, transport. And the big issue there was compulsory seatbelt wearing, they just brought in the law saying that every driver and front seat passenger had to wear seatbelt, some exceptions. This was very controversial at the time, because the nanny state and what's the government doing telling us to do this and that and the other said the government said right, we will do it for three year trial period and see what happens. And the confident expectation results were so clear cut that nobody would argue. Sure enough, it was absolutely clear that there was a reduction in deaths and serious injuries of about 20% for drivers and 25% for front seat passengers that people argued, they said they've changed the definition for serious injury so that they're saying that people are no longer seriously injured when previously said we would have been. So we said, well, let's just look at deaths. Maybe there has been an increase in rear seat deaths, or pedestrians being hit by drivers or cyclists being hit by drivers. There's actually somebody from UCL not not a statistician, geographer, called John Adams. And he had the theory that you can never improve things by improving safety precautions because people just get more complacent and careless and call that risk homeostasis. And have quite worked out how people could be just enough careless to make no change to anyway, that just wasn't true. Yes, there was a small increase in bicycle casualties with no increase in pedestrian casualties. But the powers that be in the Department of Transport got worried. And they said, well, let's have an external review by somebody very eminent. And they got the then president of the International Statistical Institute's James Durbin to do a review. And he had an assistant Andrew Harvey, who later became a professor of econometrics at Cambridge, post great postdoc student, Simon, and after a while, he got a permanent job as a lecturer somewhere around Manchester. So James Durbin asked to join the transport to load known somebody to them to do the nitty gritty work that Simon had been doing. So I spent six months at LSE. As a result, I am on their books as a former postdoctoral students at LSE would have started here. And of course, I don't have a PhD. But anyway, yeah, it was absolutely conclusive that seatbelt wearing was a good thing.
Nathan Green 10:34
Yeah, I mean, it seems odd nowadays for such a push back. The comparison I could think of is with a cyclist wearing helmets. Yeah, where it's clear, obviously, it's much more dangerous driving a speeding car without a seatbelt. It's clear this geographer was talking about this thing. Like, if you were to wear a seatbelt, you would take more risks, would you or you drive more recklessly or something like that? Right. I see. And you're saying that like you'd really have to be on the threshold for that not to be beneficial. Okay. Yeah. It's very interesting. It's, it seems like I said, it seems very strange nowadays. I remember going to a talk by one of the statisticians involved with showing how it was a bad thing to smoke on aeroplanes, when nowadays, if you can imagine and that was only stopped relatively recently. And you can imagine how strange it would be for someone to see someone smoking on a plane. Things change. So you've already been through working on very different types of projects in different groups.
Michael Baxter 11:48
Its one of the great things about being a government statistician, you never know what you are going to do next.
Nathan Green 11:52
Yeah. Well, generally a statistician. Yeah, I can speak to that. But we even within one organisation that you had these opportunities, that sounds really cool.
Michael Baxter 12:01
And so I was 27, when I started in road accidents. It's a rather interesting episode that the government is proposing to turn the clocks forward an hour so, the same time as Central Europe. And one of the arguments was that it would reduce road accidents, and I did an analysis. I found that it wouldn't. And that pretty much weakens the case for doing this change. And of course, it was never done. And that was something that's been party political. But while I was there, the GLC brought in something called Fair's Fair, where there was a massive reduction in the tube fares and bus fares. And some boroughs didn't like it was because they had to pay for it. And they brought in a court order. And the judge said there hadn't been enough public consultation. So he terminated it, and later after the proper consultation was reintroduced. But I was asked to look had it affected road accidents, because if more people are using public transport will it reduce traffic, and that should reduce road accidents. And sure enough, road accidents dropped whilst Fairs Fare was in. And rose again, when it ended, I was going to publish this in Road Accidents Great Britain. And the policy people said you can't publish this. It's against government policy. So the Director of Statistics said how can the facts be against government policy. You can stop me the publishing Mr. Vaxes commentary, you can't stop me publishing these figures. This will appear as table number one in Road Accidents of Great Britain and people will say what's this table doing here? And I'll tell them, so they backed down.
Nathan Green 12:38
Right, so that is published. So is that some sort of sweet spot of some point people will still keep using their cars and not want to use public transport?
Michael Baxter 13:36
Even if public transports was made free for people, some people are using their cars. But you know, the issue of public transport and how to improve road safety by reducing traffic issues is still a live issue about ULEZ being extended to outer London it's not so much a traffic, road traffic accident issue or if they can get elderly cars off the road, it probably would reduce road accidents.
Nathan Green 14:15
So is there someone is there a civil servant somewhere looking at that sort of thing right now?
Michael Baxter 14:20
I don't know. It wouldn't surprise me.
Nathan Green 14:22
Yeah. It's like it from what you say. It's really, really interesting. Fair's fair is called right that's fare F A R E S. Fair. F A I R. Okay, so that was to do with seatbelts and road traffic accidents. And then you were looking at bus statistics, is that right? What were you doing there?
Michael Baxter 14:45
Well, bus statistics had been a bit of a backwater, because nobody's really interested in them but I arrived at a time when something had become interesting, because the government had a policy of selling off the national bus company and the Scottish bus group, and deregulating. You can get healthy competition into it. And Minister Michael Portillo , which is his first Ministerial job, so he was keen to make an impact. So, I was doing the quarterly and annual survey of bus operators. And I found that the use of buses was going down as a result of these measures well, at the same time as these measures are introduced, correlation doesn't mean causation that this was worried people because that was contrary to government policy. And somebody else not me, was actually given the job of trying to, quote demonstrate that it was having to do with government's measures, which of course, was an impossible job, because all the other factors were slowly changing over time, whereas this was one sudden shock. So we had to be very careful what we said. And one of the issues was that the government gave money to bus operators to give free or reduced fares to the elderly. And general subsidy was route support. The government tried to say, look how much reducing this would support by giving them money, because giving them money to have reduced rates for children, which had hitherto not been specifically subsidised. So in every statistical publication had to say very carefully, that there's so much subsidy for the elderly, so much subsidy for children, which previously would include in the other group didn't please the policy people, but they have to be statistically rigorous. So that is definitely the most political job I ever had.
Nathan Green 16:31
I can imagine where politics crosses over with statistics or the facts, there's different pressures arent there.
Michael Baxter 16:37
It is always difficult if you're a government statistician. On one hand, you have to be ethical. And on the other hand, the job too much. Yes. When they brought in the community charge, the poll tax, they called it, the government decreed that it wasn't a tax, and the national accounts be reciprocal. It is technically a tax under the usual rules, but they weren't allowed to treat it as a tax or national accounts purposes. That would now be impossible, because we follow the European system of accounts. And if we suddenly started diverging from it, it would say, why are you doing that. Then I became a grade seven, and went into central statistical office, I was actually the last person to join the central statistical office while I was still part of the Cabinet Office. A couple of weeks later, it was hived off and made much bigger because various other bits of the government's statistical service put into it to make it a viable separate department. And my first job was one called title of head of methodology for the government statistical service, the idea was that I was supposed to be a source of advice to anybody who needed more technical backup. And I made a number of technical innovations, too technical to go into here. It was a very enjoyable job. For the first time in my career, I was really, really using my MSc in some sophisticated statistics, then I got moved over into the retail price index there had been an error in the RPI, the RPI is never never revised, for legal reasons. So when they had to admit a mistake, in any other statistic, they'd say, we have revised the figures that we had to say sorry, there's a mistake. As a result, these National Audit Office investigation, and they might have a number of recommendations. And my place was originally intended to implement these recommendations, which I did. But then in America, they published something known as the Boskin report, which said that the American CPI overestimated inflation by 1.2% a year and people all around the world in order to measure inflation and estimate inflation said wouldn't that be a good thing to reduce inflation at a stroke by changing the methodology and a number battles with the Treasury, we said, we're not going to change the methodology. We're not the United States, our methodology is different, it's much better actually, so we're not going to copy what they do. And then also because of monetary union in the EU, they decided to create a harmonised index of consumer prices that were produced on exactly the same methodology in every country so that they could rigorously compare inflation rates. I wasn't actually the person who went to Brussels to discuss it, but and one of the big things must be said that to combine price quotes for a particular item, you should take a geometric mean of the price changes and not an arithmetic mean as doing. Of course, the geometric means was always less than the arithmetic mean. So Treasury was saying why don't use the geometric mean in the retail price index. And we said no. The other thing about the retail price index is it was the only statistic published by the surgeon statistical office where the final decision did not rest with us. The Chancellor had certain powers to control the RBI. And the Treasury stretched that to the breaking points to interfere with what we were doing. That we were absolutely thrilled we were not going to switch to rejigger geometric mean. So after I left, the Treasury very carefully got rounded by saying well, we won't use the RPI anymore. We'll use the parliament Price Index, the CPI. Typically the CPI shows inflation rate about 1%, lower than the RBI. And at the level of information inflation, so high differences much greater. And they boasted, isn't it a good idea to use this European measure which is an international standard. So we can't be, it can't be rigged. They didn't point out that a, it hadn't been created to be used for the same purposes as the RPI, like operating pensions, or basis for negotiations with pay, and B., no other major country in the EU used it for domestic purposes, once the small ones did, because they were introduced to indices that France, Germany, Italy, carried on using their own measures will completely rub one out among the major countries. And the statistical society didn't like the CPI. And they work and be the chosen member that produced a critique of it. And something we're still within the roster, civil society taking an active interest in. And I'm now a member of the National Statistics advisory group within the rules of civil society. And of course, whenever anything to the RPi comes up. That's my opinion,
Nathan Green 21:12
Was this not settled a long time ago, people trying to interfere or change how it's calculated, or what's the current situation?
Michael Baxter 21:20
We can't change the way the CPI is calculated very easily. What we can do is say don't use it for domestic purposes, like operating pensions, produce a different measure is something that the household cost index, which is still an experimental stage, which is exactly that it's a different measure of inflation. One of the interesting things about it is that it's going to be produced for different income ranges. So you can see, it is well known that inflation affects people's incomes to buy different amounts. For example, at the moment, food inflation is much higher than inflation in most other items. And poor people spend the high proportion of expenditure on food, than rich people do. So at the moment, inflation is higher for lower income households. And this will be shown very clearly by having separate household cost indices for different groups.
Nathan Green 22:10
Right. Well, that's very topical right now. So when they talk about on the news, perhaps they should be talking about it in a bit more of a nuanced way.
Michael Baxter 22:18
You just can't on the news, you might be able to in a half hour programme devoted to the subject.
Nathan Green 22:24
Yeah, yeah. Fair enough. I think you're right. So you're still working on the RPI. So for how long were you the head of methodology for?
Michael Baxter 22:34
The Head of methodology was four years 89 to 93, then RPI was five years 93 to 98. And then I got into another controversial area in national accounts. What is the outputs of the government? If you're a manufacturer, it's easy enough, how many widgets do you produce? The government, the traditional way worldwide was to say, well, the output equals the input. You say, How many staff have you got what's your procurement, and then adjusted for inflation, and considerable work going on in many countries about measuring output more directly in Australia, they take is seriouly that people have done it full time, I was working on it part time, in addition to actually producing the routine figures, I have worked with an assistant, I think I made the correct good progress in the circumstances. And I had the interesting experience of delivering a talk to the OECD, simultaneous translation, I tabled the paper in English and in French, and delivered it in English and has been translated into several different languages as I spoke, and that was put in for fire brigades and magistrates courts, fire brigades most of their work isn't actually dealing with fires, they do things like if it's major road accident, they have to cut the people out of the cars. If there's an oil spill, they have to wash the road down to get rid of the oil. And also they spend enormous amount of time going out, giving talks and doing inspections to improve safety and prevention measures and care calls.
Nathan Green 24:03
Cats up trees?
Michael Baxter 24:05
Yes, it not a huge part of it. But the accouting is a very arcane. I didn't know if it really understood everything that goes on the National Council. The government sector just seems and I was working on
Nathan Green 24:18
Did you work in health as well did you? Was that around that time?
Michael Baxter 24:22
Health was a bit later, I left national accounts, because I was told it would be good for my career if I did a completely non statistical job, so I spent two years as head of farm animal welfare. And that was quite a nightmare of a job because there was the time we had the Foot and Mouth epidemic.
Nathan Green 24:37
Good timing
Michael Baxter 24:39
It's not glamorous and the sort of things most statisticians would never come across, like how to get Statutory Instruments through Parliament. I got three through, and what about the welfare of battery hens, and what about the welfare of intensively reared pigs, which are implementing European Union directives? So in retrospect, it was quite interesting but I didn't enjoy it at the time.
Nathan Green 25:00
Yes, okay, it's good to look back on.
Michael Baxter 25:02
And then I went to transport, that was mainly concerned with the accessibility of services. There's a thing called the social exclusion unit. And they said that maybe people are being socially excluded because they can't easily access services. For example, if you're in a system where you leave school at 16, you have to leave school at 16 and go to a further education college to do A levels. And the transport is such that you can't easily get to another education college, you dropout. If it's very difficult to get to hospital, you might miss your appointments, which will affect your health. There's a lot of issues about that. So I came up with a series of measures to estimate how easy it was to access all these different services in different parts of the country. And it was put into the Index of Multiple Deprivation, which didn't make me very popular with other people working on that subject because Accessibility Services is actually strongly negatively correlated.
Nathan Green 25:57
I suppose the difference between being public transport and being sort of connectivity, because you might have your own car. Yeah, that is interesting.
Michael Baxter 26:45
As well, access to cars, you say, well, 90% of households in this area have the car they're fine. Trouble is that the husband takes the car off to work every day, and the wife is left without a car, assuming she's got a driving licence. So, analysis of cars is actually quite a non straightforward business.
Nathan Green 27:02
Okay. It's never straightforward, is it? Probably, a rule of thumb.
Michael Baxter 27:10
Then I went to health. And I was concerned with payments to pharmacists, not the statistics of what drugs were dispensed. Somebody else doing that. But pharmacists, then some quite some years ago, paid about 2 billion pounds a year to dispense prescriptions, which can be significantly more than that now. And I had to make forecasts of demand for prescriptions. So that they could other things that we paid pharmacists for, so they could tweak the amounts to keep within budget. We're constantly negotiating with PSMC, the pharmaceutical services negotiating committee, and a substantial proportion of this 2 billion was a hidden subsidy, that pharmacists would dispense. And maybe some, and they would charge not only their fees for dispensing, but also across the cost of the medicine. And they will reimbursed at the listed price of the medicine, which often they could get it cheaper, something like 300 million, about 2 billion was what we estimated they were able to get by doing that. And this is totally above board, we knew what they were doing.
Nathan Green 28:16
Like a builder.
Michael Baxter 28:18
And it had two advantages. One is that it makes the reimbursement much simpler. We didn't have to ask for invoices. And the other was because they kept hunting around hunting around to get lower prices. It kept prices down. So it was actually a good system. And I did a survey of pharmacists to find out how much money they're actually making. And I think it was something like 600 million rather than 300 million. And because the pharmacist didn't like that, because we said that we'll we'll cut the reimbursement prices. And they said pharmacists can go bankrupt if you do it. And they challenge my analysis. And they got a consultant who said no I'd done it all wrong. And thus told my bosses, they've got a consultant, we've got to get a better consultants. So I asked people at the ONS, who do you go to for advice on surveys? Southampton University. So I approached Southampton. And they recommended Pedros De Silva, who was later actually the president of the International Statistical Institute very prominent statistician. And he looked at what I've done, he said on the whole its right, the sampling error is a bit better, bit bigger than you've estimated otherwise, your central estimate is correct. So pharmacists kept arguing that basically the project would cut the prices, and the next year I did another survey and found they're still making too much money so we cut the prices again. So in the context of the total cost of the NHS, it's not enormous. It's appreciable.
Nathan Green 29:44
Yes, and if it's it sounds like though rigorous analysis. Its totally justifiable.
Michael Baxter 29:50
And after that, for various reasons, I left the civil service and I ended up in market research. We work on how many people watch television, which is a big deal because
Nathan Green 30:00
Advertising is it?
Michael Baxter 30:03
Huge fortunes are spent on television advertising. And it's not we don't just do it in this country, we do it everywhere from the Philippines to Peru.
Nathan Green 30:10
Wow, okay. Didn't they used to have a box like that you'd have on top of the television to do that. And so how are they doing nowadays?
Michael Baxter 30:20
With classical television actually thats what you do, which, of course, it's not as simple as that you'd like to have a panel and make sure the panel is unbiased. And so on. Things are getting more and more complicated. In 1990, there were only four television channels in this case
Nathan Green 30:38
I remember Yes.
Michael Baxter 30:41
We monitor hundreds, literally hundreds now.
Nathan Green 30:44
Are there enough viewers across all those 100 channels in order to monitor them all? You know, like, the 90% of people watch 10 channels only.
Michael Baxter 30:55
Small channels that are so low, we can give you an annual estimate. Whereas for the BBC and ITV, we can give you minute by minute. The people who own these small channels want to know how many people are watching them.
Nathan Green 31:09
Yes. I suppose so. Yes. But you're still involved with the Royal Statistical Society?
Michael Baxter 31:14
Oh, yes. As I said, I'm a member of the National Statistics advisory group. And I've just helped to set up a new section within the group on financial and economic statistics.
Nathan Green 31:26
What are you going to cover with that group?
Michael Baxter 31:29
Oh, quite a lot. Insurance. National Accounts. Anything thats kind of our financial economic.
Nathan Green 31:36
Yeah, yeah. Well, it seems like a group that should have been setup before now rather than recently.
Michael Baxter 31:43
I should have said, I was chairman of a different section of the RSS at one point. The official statistics section.
Nathan Green 31:51
Yeah, this is arguably the oldest type of statistics there is right.
Michael Baxter 31:57
It was one of the first sections that was set up.
Nathan Green 31:59
Yes, yes. Now, I've been involved in the medical secction for a while. That's also a very old one. Not these Data Science, Sports analytics people, the new crowd. Right. I mean, the things that you've worked on the variety, the interest is remarkable. So if you were something that you wish you'd known when you were a student at UCL that you know now or a word of advice to graduating students, what do you think that would be?
Michael Baxter 32:29
I'm not sure the government's statistical service is as good a career now as it was then. And I think if I were a new graduate now, I'd join the government's statistical service.
Nathan Green 32:41
What not to do. I mean, I think what you've recounted about your career it tells its own story in that light with a statistics degree that is absolutely massive variety of things that you can apply yourself to certainly different interesting problems that the same skills can the different contexts and different types of problems. So I think that speaks volumes that speak for itself. So I'm gonna end it there, Michael, and to say, thank you very much for speaking to us. It's been absolutely fascinating. I really enjoyed speaking to you, and I wish you well in your current work and whatever other completely different interesting work you do after that. So thank you, Michael.
Michael Baxter 33:22
Thank you.
UCL Minds 33:25
UCL minds brings together the knowledge, insights and ideas of our community through a wide range of events and activities that are open to everyone.