Lunch with the FT - Andrew Dilnot

The economist, broadcaster and Oxford principal explains why only one statistical question has ever troubled him: how many penguins are there in Antarctica? By Tim Harford

As I enter the porter's lodge at St Hugh's College, Oxford, I fleetingly reflect that I may be about to receive an intimidating tutorial from the college principal, Andrew Dilnot. It is not that the economist has a stern reputation, but today's circumstances are unusual. He recently stepped down from presenting More or Less, a BBC Radio 4 series about numbers in the news. I have been recruited as the new presenter, and am ready to be patronised - or worse.

I needn't have worried: as he strides into the lodge in a pale brown linen suit and blue tie, Dilnot's smile is genuine enough, and as we walk together through north Oxford's leafy residential streets, he is more eager to identify shared acquaintances in the world of economics than to lecture me on the art of radio presenting.

Dilnot has suggested we visit the Cherwell Boathouse restaurant. It's tucked away down an alleyway, just next to a shed where punting poles, oars and cushions are hired out. On a sunny day in Trinity term, the slow-drifting river would be packed with jostling punts full of rowdy students. Today we gaze out on to a silent river and a row of moored boats.

Once we're seated, I again fear that my tutorial may begin. But despite his formidable qualifications (CBE for ''services to economics''; pro-vice chancellor of Oxford) and 11 years as director of the Institute for Fiscal Studies - the UK's leading tax think-tank - Dilnot is not prone to lecturing. He preaches a message of statistical self-help, both through More or Less, and his new book, The Tiger That Isn't, written with the radio series' creator Michael Blastland - it's already into its third reprint.

''We are trying to show people how they can interpret the numbers that are thrown at them,'' Dilnot says. That is often a less technical business than it seems. He advises people to ask simple questions, such as: ''Is that a big number?'' The answer is often not hard to work out. One of Dilnot's favourite examples is the Labour government's promise in 1997 to spend £300m over five years to create a million new childcare places. But if you take apart the numbers it looks less impressive: £300m for a million places is £300 per place. Over five years, that's £60 per place - or a little over a pound a week. Good luck finding childcare for that sort of money.

We pause for a moment to contemplate the menu. The waitress tells us the foie gras is off. Dilnot declares he doesn't eat foie gras, anyway, having once visited a farm in France and seen how it was made. Even without the foie gras, the choices all look a little rich, until Dilnot points me towards the lighter set menu page tucked away at the back of the large leather folder. His belief in self-reliance then reasserts itself, as he refuses to offer me any recommendations.

He plumps for the Tivoli ham with asparagus, followed by grilled chicken and vegetables. I am still wondering what Dilnot saw at the farm, and choose a fried lentil cake, with gazpacho to start. Dilnot refuses a glass of wine, claiming it will send him to sleep. He also turns down the offer of bread. ''I am eating for the college at the moment,'' he says, patting a nonexistent pot belly. That leads us on to the evidence for and against different diets, but Dilnot is more taken by the research into links between diet and crime.

''I can't remember if we mention it in the book,'' he says (they do), ''but there was a randomised trial of prisoners in the late 1990s. Some got nutritional supplements and others did not.'' The trial showed big improvements in the behaviour of prisoners who received the supplements. It's an example of how to do statistical work properly, and it contrasts with most of the crimes against counting that Dilnot and Blastland tend to document. But Dilnot complains that the results have been binned by the Home Office, and that the Home Office refuses to explain why there has been no follow-up work.

Our starters arrive, and we stop the professional chat to eat. The gazpacho tastes good, and Dilnot professes to enjoy his ham. But neither dish is exciting enough to divert us from the conversation. I ask Dilnot whether the failure to act on good research, such as the prison experiment, suggests an inbuilt bias somewhere in favour of sloppy statistics over careful work?

''There should be a market for good statistical work,'' he says. ''But we are biased in favour of a dramatic story.'' He relates one example: rigorous statistical work on the safety record of UK railways after privatisation showed that, despite popular belief, privatisation accelerated a pre-existing trend towards ever safer rail travel. But while Dilnot discussed this work in depth on his radio programme, it didn't get much interest in the rest of the media. ''Senior journalists told us that Hatfield [the 2000 train derailment that killed four people] had proved that things had got worse.''

But perhaps, I suggest, the fault lies with the statisticians themselves, for being uninterested in - or incapable of - presenting their research in a way that is engaging, persuasive and comprehensible to the rest of us? He acknowledges this, ruefully, but also points out how easy it now is for anyone to get online access to good statistics and find out for themselves what is happening.

''It doesn't need to be complicated. Often you just need the right number.'' But not enough people seem interested in doing that. Dilnot is perplexed by this. ''I love looking at the numbers. There's always a surprise there. When I'm researching a new topic, I get a boyish surge of excitement whenever I discover another thing I didn't know.'' But this is probably not surprising, coming from a man whose earliest professional memory is of analysing the government's Family Expenditure Survey using a computer that spat out numbers in base 16.

We can't all give numbers the sort of loving embrace that Dilnot does. But we can, he believes, do a lot better than we think. One simple trick is to try to humanise statistics. Faced with a question such as: ''how many petrol stations are there in the UK?'', ask yourself how many petrol stations there are in your town, and how many people. It's the first step towards grasping a sensible answer to the bigger question.

As we tuck into our main courses, I can't resist teasing him about this bigger game of humanising statistics. Dilnot has written that just about the only question that can't be partially answered with reference to personal experience is: ''how many penguins are there in Antarctica?'' I ask him if he actually knows the answer; two million, he suggests, before launching into an extended monologue about the difficulties of a credible penguin census.

I steer the conversation away from penguins and back to something in his book that intrigues me. The book alludes to what the sociologist Joel Best labelled ''the worst social statistic ever''. I ask him to tell me more. ''The worst social statistic ever,'' Dilnot recalls, ''is that the number of murders in the US has doubled every year since 1953.'' He leaves it to me to figure out why it's so bad. It isn't hard: even if there had been just one murder in 1953, I reckon aloud that the entire population of the US would have been dead sometime in the 1970s. (Later, I check my numbers, and I'm not far off. There would have been more than 30 million murders in 1978, almost 70 million in 1979, and the last available victim would have perished in 1980.)

Over a decaffeinated coffee (he resists dessert but easily persuades me to have a strawberry creme brulee), Dilnot jokes about how deeply he absorbed the details of the UK economy, after his many years at the Institute for Fiscal Studies (IFS). ''My wife said that if I left, I'd only be useful for Mastermind, with my specialist subject as UK Budgets of the 1980s and 1990s.'' In the end, he left in 2002, for a much broader remit at St Hugh's. Why did he make the move? There's the commute, he admits. The IFS is in London, but ever since Dilnot went to Oxford as a student, he has lived in the city, now with a wife and two daughters.

But there's more to it than that. Dilnot says that he came to realise that one of the IFS's most valuable - if unmeasured - outputs was of good economists. (The Institute's alumni include the economics editors of the BBC and of The Financial Times.) An Oxford college would, he reasoned, be an even more important training ground for future high-flyers. And he is passionate about the way the college system brings together academics from different disciplines, who would normally never have a reason to talk to each other.

Given Dilnot's familiarity with how statistics are produced - like sausages, the process is not pretty up close - it is surprising that he still believes that numbers can do a lot of good. He delights in skewering the perverse consequences of the New Labour fad for using targets to try to improve the performance of the public sector. And yet he says he feels torn: ''Despite all the unforeseen and adverse side effects, there seem to have been important improvements in performance, too.''

I ask him how government targets could be set more sensibly. He argues for simplicity, surprise audits, sanctions for those who ''game'' the system, and a willingness to change frequently what is measured. All these recommendations are aimed at persuading the managers of schools and hospitals that the best way to win plaudits is to do a good job.

The time comes to settle the bill and leave, and we walk back to the college together. He gives me some (solicited) advice about how to make a radio series, and then asks me what I plan to do with my afternoon. I tell him I am going to find a quiet place to write up this interview, so he offers the college library. ''Somehow the smell of Oxford libraries puts me to sleep,'' he says. I always found that too, so it is refreshing to hear one of the university's senior figures agreeing. I head to the library feeling relaxed - and far from intimidated.

Tim Harford is the author of ''The Undercover Economist'' and presenter of BBC Radio 4's ''More or Less''.

当我走进牛津大学圣休学院(St Hugh's College)的门房时，我脑中突然闪过一个念头：该学院的院长迪尔洛特可能将给我上一堂令人胆怯的辅导课。这并不是说这位经济学家有着严厉苛刻的名声，而是今天的情况有些特殊。他最近离开了《多多少少》(More or Less)的主持工作，这是英国广播电台(BBC)4台一个关于新闻里的数字的系列节目。我被聘为新主持人，所以我已做好了接受说教（甚至是更糟）的准备。

我完全是杞人忧天：当他跨进房间时，他笑得很诚恳。他穿着一身淡褐色亚麻西服，戴着蓝色领带。当我们走在牛津住宅区的林荫道上时，他更愿意谈论我们都认识的经济学界人物，而不是就电台主持艺术对我进行说教。

河滨小餐厅

迪尔洛特建议我们去查维尔船屋(Cherwell Boathouse)餐厅。它藏匿于一条小巷里，紧挨着一个出租船竿、船桨及坐垫的小屋。在夏季学期阳光明媚的日子里，这条舒缓的河流上会挤满了相互碰撞的平底船，里面坐着大呼小叫的学生们。我们今天只看到一条寂静的河流，平底船在岸边被固定成一排。

在我们落座以后，我再次担心辅导课即将开始。但是尽管他的资历令人敬畏——因“对经济学的贡献”而获得大英帝国司令勋章(CBE)，还是牛津大学副校长——并担任英国主要税务智囊机构伦敦财政研究所(Institute for Fiscal Studies)所长一职长达11年，他却不是那种有说教倾向的人。通过《多多少少》，以及他与该节目创作者迈克尔•布拉斯兰德(Michael Blastland)合著的新书《数字不是老虎》(The Tiger That Isn't)，他赞成人们应在统计方面自己动脑筋。《数字不是老虎》一书已经第三次再版。

他表示：“我们试着告诉人们怎样去诠释碰到的数字。”通常，这并不像表面看起来那么复杂。他建议人们问一些简单的问题，例如：“这是一个大数吗？”答案往往并不难算出。迪尔洛特最喜欢的一个例子是英国工党政府在1997年承诺，在之后5年里支出3亿英镑来增加100万个托儿位置。但如果你将这个数字分解，它看起来就没什么了不起的了：100万个位置共3亿英镑，合每个位置300英镑。分摊到五年就是（每年）每个位置60英镑，相当于每周1镑多点。谁要用这么点钱去找托儿服务，祝他好运了。

我们停了一会儿来考虑点什么菜。女招待告诉我们鹅肝没有了。迪尔洛特声明，自从他去过一家法国农场并看到了鹅肝的制作过程后，他就再也不吃鹅肝了。就算没有鹅肝，可选的菜品也有点太多了，直到迪尔洛特为我指出了藏在大皮夹子后面的比较清淡的套餐页。他拒绝向我做任何推荐，主张自己动脑筋的立场再度显现出来。

他选了Tivoli火腿配芦笋，及烤鸡配蔬菜。我一边想迪尔洛特在农场看到了什么，一边点了一个油炸小扁豆糕，头盘是西班牙凉菜汤。迪尔洛特拒绝了喝一杯葡萄酒的建议，声称这会让他想睡觉。他也没有吃面包。“我现在是在为学校进食，”他拍着一个并不存在的啤酒肚说道。这将话题引向了不同食谱优缺点的证据，但迪尔洛特对饮食和犯罪之间的关系的调查更感兴趣。

获得营养补充品的犯人有很大改善

“我记不起来我们是否在书中提到了这一点，”他表示（他们提到了），“在上世纪90年代末，一个实验随机挑选了一些犯人，其中有些得到了营养补充品，有些没有。”实验显示，那些获得了营养补充品的犯人在行为方面有了很大的改善。这是一个表明如何正确进行统计工作的范例，与迪尔洛特和布拉斯兰德倾向于记录的很多在计算上犯下的“罪行”形成鲜明对比。但迪尔洛特抱怨称，英国内政部(Home Office)将这个结论抛到了一边，并拒绝解释为什么没有后续研究。

头盘上来时，我们停下了专业方面谈话，开始进餐。西班牙凉菜汤的味道很不错，迪尔洛特也表示自己的火腿很不错。但这两盘菜的美味程度都还不足以将我们的注意力从谈话引开。我问迪尔洛特，上述监狱实验等优秀的调查未能令政府采取行动是否表明，某些地方存在着更倾向于马虎统计而非精心调查的内在偏好。

“好的统计工作应该是有市场的，”他表示，“但我们的偏好倾向于引人注目的故事。”他以英国铁路私有化之后的安全记录为例，严谨的统计调查显示，与流行的看法相左，私有化加快了此前已经存在的铁路出行越来越安全的趋势。但当迪尔洛特在自己的广播节目上对此项调查进行深入讨论时，并没有引起其它媒体的太多兴趣。“资深记者们告诉我们，哈特菲尔德（2000年的火车出轨事件，导致4人死亡）证明，铁路安全状况有所下降。”

但是我提出，问题可能出在统计学家自己身上，因为他们没有兴趣或没有能力以一种吸人入胜、说服力强和易于理解的方式将自己的调查结果展示给其他人。他遗憾地承认了这一点，但又指出，现在任何人都可以很容易地在网上找到好的统计数据，自己搞清楚事情的真相。

“这并不需要很复杂。你一般只需要得到正确的数据。”但没有多少人对此抱有兴趣。迪尔洛特对此表示不解。“我喜欢看数字。那里总有一些惊喜。当我研究一个新客题时，每当我会发现自己从前不了解的东西时，我都会感到一股孩子般的兴奋涌上心头。”但这可能没有什么令人吃惊的，毕竟它来自一位事业之初曾用输出16进制数字的计算机为政府分析家庭支出调查的人。

我们无法都像迪尔洛特那样热情地投身于数字当中。但他相信，我们可以比自己想像的做得好得多。一个简单的技巧就是将统计数据人性化。当面对“英国有多少家加油站？”这样的问题时，问问你自己你的镇上有多少家加油站，镇上人口又是多少。这是得出更大问题的一个合理答案的第一步。

当我们吃主菜时，我忍不住就将统计数据人性化这个更大的问题去揶揄他。迪尔洛特曾写到，唯一一个无法凭借个人经验得到部分答案的问题是：“南极洲有多少只企鹅？”我问他是否真的知道答案，他提出有200万，随后就进行可信的企鹅人口调查的困难开始了长篇大论。

我将谈话从企鹅的话题引回到他书中引起我兴趣的一些东西。书中提到了社会学家约尔·贝斯特(Joel Best)称之为“有史以来最差的社会统计”。迪尔洛特回忆称：“有史以来最差的社会统计是指自1953年以来美国的谋杀案数量每年翻一番。”他让我自己去搞清楚这个数据为什么如此之差。这并不难：就算1953年只有一桩谋杀案，我估计整个美国人口在上世纪70年代的某个时间就应该死光了。（后来，我检查了自己的估算，基本上差的不多。根据假设，1978年会有3000万桩谋杀案，1979年近7000万，最后一个受害者会在1980年死亡。）

在喝无咖啡因咖啡时（他拒绝吃甜点，但轻易地说服我点了一个法式草莓焦糖布丁），迪尔洛特开玩笑称，在财政研究所工作多年之后，他已经对英国经济了如指掌。“我妻子说如果我辞职离开，我就只能去参加Mastermind(译者注：英国电视上的一个智力问答节目，参赛者可以挑选自己精专的区域进行挑战)了，专长项目是上世纪80到90年代的英国预算。”最后，他在2002年离开了财政研究所，到圣休学院干起了一份责任更加广泛的工作。为什么他要换工作呢？他承认，上下班路程太长。财政研究所在伦敦，而迪尔洛特自牛津的学生时代起，就一直住在牛津。现在他和妻子与两个女儿生活在一起。

但这并非唯一的原因。迪尔洛特表示，他意识到财政研究所最宝贵（虽然无法衡量）的产品是优秀的经济学家。（BBC和英国《金融时报》的经济编辑曾在该研究所就职。）他考虑，在培养未来精英方面，牛津的一个学院会是更重要的训练场。此外，他非常喜欢大学系统可以将不同领域的学者聚集到一起这一点，否则，这些人永远都不会与彼此交谈。

考虑到迪尔洛特对统计数据的产生过程非常熟悉——就像香肠一样，在近处观察，这一过程并不美观——他仍然相信数字可以做很多贡献，这令人惊讶。他喜欢将新工党(Labour)热衷于尝试利用指标来改变公共部门业绩的风气的反面结果串连在一起。但他却说，他有种矛盾的感觉：“尽管存在那些不可预料及反面的副作用，业绩似乎有了不小的提升。”

我问他，政府如何才能将指标定在更为合理的水平上。他表明重点在于简单易行、突击审计、惩罚那些对系统“进行博弈”的人，以及愿意常常改变衡量的对象。所有这些建议的目的都是为了让学校和医院的管理者相信，赢得喝彩的最佳方法就是做好工作。

最后到了结账离开的时候，我们一起走回学院。他为我提供了（经过我的一番恳求）一些关于如何做好电台节目的建议，然后问我下午打算干什么。我告诉他自己要找一个安静点的地方写下这篇采访，于是他提议我去学院图书馆。他表示：“不知什么原因，牛津大学图书馆有种让我昏昏欲睡的气味。”我也一直有同感，所以听到一位大学高层人物这么说真的让人耳目一新。我在走向图书馆的路上感觉很放松，一点也不胆怯。

提姆•哈福德(Tim Harford)是《卧底经济学家》专栏作家，英国广播电台4台《多多少少》(More or Less)节目的主持人

• 6小时掌握学英语的秘诀！——点击看答案
• 专治男女记不住单词的秘密