Tuesday, 30 April 2013

Great British Class Survey

At the beginning of April, a study entitled A New Model of Social Class: Findings from the BBC’s Great British Class Survey Experiment was published. It received huge media attention in the UK. Indeed, it was picked up by The Guardian, The Telegraph and the Daily Mail, as well as the BBC (who helped collect data for it).

In this excellent blog post, Colin Mills debunks the study's methodology and findings.

Thursday, 25 April 2013

Can standardised tests measure creativity?

Despite their widespread use in education (particularly in the US), standardised tests are often criticised for being too narrowly focussed. As the education theoriest William Ayers argued, "Standardized tests can’t measure initiative, creativity, imagination, conceptual thinking, curiosity, effort, irony, judgment, commitment, nuance, good will, ethical reflection, or a host of other valuable dispositions and attributes."

Many of the traits Ayers refers to are clearly very difficult to measure––some would say impossible. However, a recent study by Park et al., entitled Ability Differences Among People Who Have Commensurate Degrees Matter for Scientific Creativity, looked to see whether standardised test scores were predictive of one particular such trait, namely scientific creativity. Their sample comprised participants from the Study of Mathematically Precocious Youth, each of whom was in the top 1% of (age-adjusted) mathematical ability in childhood. They followed these individuals' careers, and looked to see whether the ones who scored highest in the math section of the SAT (SAT-M) at age 13 had tended to be more creative in adulthood. Creativity was measured in several ways: number of scientific publications in peer-reviewed journals, number of humanities publications in peer-reviewed journals, number of patents, and number of patents for Fortune 500 companies.

As to publications in peer-reviewed journals, they found (as shown below) that, for both master's and doctoral degree holders, individuals who scored in the top quartile of the group's SAT-M scores were significantly more likely to have had at least one scientific publication than those who scored in the bottom quartile. Unsurprisingly, there were no significant differences between the top and bottom quartiles in the proportion having had at least one humanities publication. As to patents, they found--once again--that those who scored in the top quartile were significantly more likely to have had at least one than those who scored in the bottom quartile. And this held for both kinds of patent and all three degree-types.

Overall, Park et al.'s findings run contrary to Ayer's claim that standardised tests cannot measure creativity. As a follow-up, it would be interesting to see whether, among children gifted in verbal reasoning, SAT-V score at age 13 is predictive of subsequent achievement in, say, journalism and the humanities.

Monday, 22 April 2013

Some healthy skepticism

"Rejecting a macroeconmic idea (Reinhart and Rogoff) over an excell error is like falsifying astrology over a computer glitch."
--Nassim Taleb

Sunday, 21 April 2013

Does smoking marijuana lower cognitive performance?

A study by Meier et al., entitled Persistent Cannabis Users Show Neurophysiological Decline from Childhood to Midlife, was published last year, at which time it received widespread media attention. But what did it actually find, and can its results be trusted?

Meier et al. followed around 1,000 individuals (comprising a single New Zealand birth cohort) for nearly 40 years, assessing their marijuana use at ages 18, 21, 26, 32 and 38, and their (age-adjusted) cognitive performance at ages 7, 9, 11, 13 and 38. According to the authors, there has only ever been one other cohort study of marijuana use and cognitive performance, namely Neurocognitive Consequences of Marijuana--A Comparison with Pre-drug Performance by Fried et al. (2004).

This older study observed marijuana users after a mean duration of usage of only 2 years, and had a sample size of only 121. Thus, Meier et al.'s study is by far the most comprehensive analysis of the subject to date. Before describing exactly what Meier et al. found, it was worth noting that--in spite of the small sample and limited duration of exposure after which users were assessed--Fried et al. documented significant differences in cognitive performance between heavy users (those smoking 5 or more joints per week) and controls. These differences (which were adjusted for prior performance, along with a number of other controls) are shown in the graph below. However, it is also important to note that Fried et al. did not find any significant differences between former users and controls.

As noted, Meier et al. assessed users' cognitive performance in childhood and then again at age 38. Their main finding was that cognitive performance between childhood and age 38 decreased in a roughly linear fashion according to the number occasions on which regular marijuana usage (4 or more joints per week) was reported. Users who never reported marijuana usage experienced a slight increase in cognitive performance between childhood and age 38, those who reported regular usage on only one occasion experienced a small decrease in cognitive performance, while those who reported regular usage on 3 or more occasions experienced a large decrease in cognitive performance. This trend is shown in the graph below (taken from the original paper); the scale on the horizontal axis is change in IQ (in standard deviation units).

Meier et al. observed the same pattern of decline on a variety of other measures of cognitive performance, and after controlling for sex and years of education. They also found a strong positive association between informant-reported memory problems (in the user) and number of episodes of regular marijuana usage. Interestingly, they observed a very small decline in cognitive performance among users who had begun smoking marijuana regularly after age 18, which suggests that regular marijuana usage may be more deleterious in adolescence than in adulthood. Finally, they observed a decline in cognitive performance even among (adolescent-onset) users who reported infrequent marijuana usage in the year before assessment at age 38, which constitutes weak evidence that cessation of usage may not yield full restoration of cognitive performance.

Despite having been released only a year ago, Meier et al.'s findings have already been called into question. In particular, Rogeberg (2013) published a paper entitled Correlations Between Cannabis Use and IQ Change in the Dunedin Cohort Are Consistent with Confounding from Socioeconomic Status. He makes the reasonable point that Meier et al.'s results are consistent with the idea that individuals predisposed to experience cognitive decline in adulthood select into marijuana usage. His conclusion is that the decline in cognitive performance among marijuana users observed by Meier et al. is almost certainly an over-estimate of the causal effect of marijuana usage.

Overall, there seems to be reasonably good evidence that smoking marijuana lowers cognitive performance in the short-term. However, the evidence that smoking marijuana permanently lowers cognitive performance is, at the present time, still relatively weak.

Wednesday, 17 April 2013

Thatcher's popularity: 1980s and today

In this post, I attempt to evaluate Thatcher's popularity, both in the 1980s and today. The graph below displays vote shares in UK elections since 1945. It confirms that Ken Livingstone is correct in pointing out that Thatcher's landslide victory in 1983 was not so much due to her being extremely popular, as to Labour being extremely unpopular. However, Livingstone's claim that "[the Conservatives'] long-term decline continued under Thatcher" is quite misleading. Arguably, both Labour and the Conservatives have been in decline relative to other parties since the February 1974 election.

What do people think of her today? Since her death last week, at least two major polls have been carried out in an attempt to answer this question: one by YouGov in partnership with The Sun (just under 2000 respondents), and one by ICM in partnership with The Guardian (just under 1000 respondents). Results from the two polls are highly similar. The YouGov poll documented that 52% of people believe she was a good or great Prime Minister, while 30% believe she was a bad or terrible Prime Minister. And The ICM poll found that 50% of people believe she was a fairly good or very good Prime Minister, while 34% believe she was a fairly bad or very bad Prime Minister.

How does Thatcher fare in comparison with other British Prime Ministers? In 2008, Newsnight asked the British public to vote for the greatest post-war Prime Minister, a request to which 27,000 people responded. Thatcher ended up in 3rd place, behind Winston Churchill and Clement Attlee. In 2007, Daily Politics asked its viewes to vote for their faveourite Prime Minister among a selection of ten (excluding Churchill). Thatcher came in 1st place with 49% of the vote; Clement Attlee came in 2nd with 32% of the vote. Finally, the 2013 YouGov poll asked respondents to name the greatest Prime Minister since 1945, and Thatcher finished in 1st place with 28% of the vote. 

Overall then, Thatcher appears to be quite popular with the British public. Of course, opinion on her is still rather polarised.

Tuesday, 16 April 2013

Dean Baker quote

"If you want to talk to someone from Goldman Sachs, call the Treasury."

Monday, 15 April 2013

Did Thatcher destroy the British coal mining industry?

Following the death of Margaret Thatcher last week, commentators on both the left and the right have been compiling lists of myths about her time in office. Here, I examine the widespread claim that she destroyed the British coal mining industry. In particular, I assemble three graphs for the period 1950-1994: one depicting the trend in coal mining output; one depicting the trend in total number of coal mines; and one depicting the trend in coal mining employment. Evidently, the decline of the British coal mining industry began in the late 1950s, and then continued through both Labour and Conservative governments up until the mid 1990s. Moreover, the decline was not unusually rapid during Thatcher's time in office. The data are from the Department of Energy and Climate Change.

Sunday, 14 April 2013

Is there such a thing as an unhealthy food?

The short answer is no. The long answer is no (with one or two exceptions).

Consider two diets: in the first one, you only eat green salad; in the second one, you only eat McDonald's hamburgers. Although I certainly wouldn't recommend the second diet, it is unquestionably healthier than the first. Protein (like fat, but unlike carbohydrate) is an essential nutrient, and green salad does not contain any. Although you might eventually catch scurvy through lack of vitamin C on the hamburger diet, you would be dead in a matter of weeks (or a couple of months at best) on the green salad diet. But how can this be? Green salad is "healthy", whereas McDonald's hamburgers are "unhealthy".

Before I explain why I don't think there's such a thing as an unhealthy food, it is necessary to define the term 'healthy'. A reasonable definition of 'healthy' (as applied to things humans do or eat) is 'conducive to the avoidance of morbidity'. In other words, one thing is healthier than another thing if it offers greater protection from infection and disease.

A category error is the assignment of a property to an object that could not meaningfully possess that property; the phrase was coined by the philosopher Gilbert Ryle. An example of a category error would be describing a carrot as intelligent, or––as in the case of Ryle's original example––conceptualising Oxford University as a single physical entity. I would argue that (in general) it is a category error to assign the property of being healthy to a particular food. I.e., it does not make sense to say, "green salad is healthy" or "hamburgers are unhealthy", just as it does not make sense to say "carrots are intelligent" or "Oxford University is a single physical entity". Comparatively, I would claim it does make sense to speak of diets as being healthy or unhealthy.

One reason why it is not meaningful to refer to a particular food as unhealthy is that, for any substance, there is always a quantity which, if ingested, would not cause any adverse consequences. For example, you would not suffer any adverse consequences if you happened to ingest 0.00000000000000000001g of plutonium. As a friend pointed out to me, this is exactly the point chemists have in mind when they say, "it's the dose that makes the poison." At the other end of the spectrum, there is always a quantity of a substance which, if ingested, would kill you. Trivially, even nutrients that are readily excreted by the body would build up and cause death if sufficiently large quantities (e.g., millions of kg) of them were ingested. More interestingly, there are rare cases of water overdoses.

A fair response to the preceding argument is that when a person describes a food as unhealthy she is usually referring to a typical portion size of that food. For example, someone might say, "it would be unhealthy for me to eat this piece of cake." Assuming it is possible to set bounds on what a typical portion size constitutes, I would argue that this response is unpersuasive.

The first reason why it is not persuasive is that, for any typical portion size (of any food), there is a frequency of ingestion of that portion size that would not lead to any adverse consequences. For example, suppose you think that it would be unhealthy to eat a piece of cake. Clearly, if you only did so once, the consequences for your health (as defined above) would be utterly negligble. The second reason why is that the effects of injesting portions of particular foods on health are not additive. In particular, the impact upon your health of eating a portion of some food depends on what else you've eaten recently. Suppose you believe that it would be unhealthy for you to eat a piece of cake. Would you avoid consuming a piece of cake on health grounds if you were starving in the desert and there was nothing else to eat?

At this point, a conceivable response is that when a person describes a food as unhealthy she is actually referring to a typical frequency of injestion of a typical portion size of that food in the context of a typical diet. In that case, I would agree, her claim is meaningful. But notice that she is now talking about diets and not individual foods. She is saying, "given what else I generally tend to eat, adding a certain number of portions of a particular food would be unhealthy relative to not doing so", which is just the same as saying "my usual diet plus a certain number of additional portions of food is less healthy than my usual diet as it stands." In other words––as stated above––it only makes sense to refer to diets as healthy or unhealthy, not individual foods.

Clearly, a diet made up entirely of McDonald's hamburgers (e.g., four quarter pounders per day) is much less healthy than a diet that includes fish, vegetables, grains, etc. However, it does not follow that as soon as you eat a McDonald's hamburger you somehow become unhealthy. There are really only two characteristics a diet needs to possess in order for it to be healthy. First, it should not be too low or too high in calories; eating 2500 calories per day is much healthier than eating 5000 calories per day (unless you're doing a great deal of excerise or you happen to be extremely tall). Second, it should be balanced. It is important to eat a variety of foods: vegetables, fruit, nuts, grains, and a range of meats; or if you're a vegetarian, a range of pulses and beans.

Before discussing one caveat to my argument, it is worth addressing what I believe is another source of confusion in the present context, namely the difference between 'unhealthy' and 'fattening'. People often refer to foods as unhealthy when what they really mean is fattening. Consider all the individuals in the developed world who are not close to being morbidly overweight. Adding a couple of portions of cake to those individuals' diets each week would not have any discernible impact on their health, but would make them a little bit fatter. In the case of fatness (as opposed to health), the effects of injesting portions of particular foods are more-or-less additive. Because of this fact, it is meaningful to describe one food as being more fattening than another. And what is meant by 'fattening' is something like 'moorish and unfilling'. For example, 500 calories of chocolate cake has very different effects on the neurological pathways governing satiety than 500 calories of boiled potatoes.

An important caveat is that there are some individuals who will suffer adverse conseqeunces from consuming even a single portion of a particular food. Such individuals may have allergies (e.g., to peanuts) or may have diseases such as diabetes, which prevent them from eating sugary foods when their blood-glucose level is high. Obviously, however, these people are quite capable of injesting very small quantities of the relevant foods.

Friday, 12 April 2013

Is gun ownership associated with homicide rate across countries?

In the wake of several tragic shootings in the US, the debate over gun control there has gained new zeal. Here, I present graphs depicting the cross-national relationship between gun ownership and homicide rate separately for: all countries with available data, and the OECD countries. In neither case is there any relationship between the two variables, contrary to some claims in the media. The data on gun ownership are from the Small Arms Survey 2007, while the data on homicides are from the United Nations Office of Drugs and Crime. Of course, just because there does not appear to be any bivariate relationship between gun ownership and homicide rate across countries, it does not mean that homicide rate is not affected by gun ownership. And indeed, a number of interesting studies have attempted to tease out the effect of gun ownership on homicide rate in the US using sophisticated econometrics.

Wisdom from P. J. O'Rourke

I stumbled across a few amusing quotes of his:

"When buying and selling are controlled by legislation, the first things to be bought and sold are the legislators."

"Always read something that will make you look good if you die in the middle of it."

"The good news is that according to the Obama administration the rich will pay for everything. The bad news is that according to the Obama administration you're rich."

"The Democrats are the party that says government will make you smarter, taller, richer and will remove the crabgrass from your lawn. The Republicans are the party that says government doesn't work, and then they get elected and prove it."

"A hat should be taken off when you meet a lady, and left off for the rest of your life."

Thursday, 11 April 2013

Should Thatcher get a ceremonial funeral?

I agree with the author of this article--the case for giving Thatcher a so-called ceremonial funeral is extremely weak.

Has social mobility declined since Thatcher became prime minister?

The author of this article about Thatcher's legacy concludes--among other things--that social mobility in Britain has declined since Thatcher became prime minister. He does so on the basis of a very widely cited paper, namely Changes in Intergenerational Mobility in Britain by Blanden et al. This study analysed data from two cohort surveys: the NCDS, which interviews all children born in Britain during a particular week in 1958; and the BCS, which interviews all children born in Britain during a particular week in 1970. It found that parents' earnings quartile was a better predictor of children's earnings quartile in the 1970 cohort than in the 1958 cohort.

Notwithstanding the popularity of this paper, its chief finding has been contradicted by subsequent research. In a paper titled Has Social Mobility in Britain Decreased? Reconciling Divergent Findings on Income and Class Mobility, Erikson and Goldthorpe point out that Blanden et al.'s study possesses a number of methodological weaknesses. First, due to poor quality data on self-employed individuals' incomes, some cases had to be excluded from the analysis. Second, Blanden et al. had to utilise one measure of parents' earnings when analysing the NCDS and another when analysing the BCS (the NCDS does not contain a variable for total family income at child's age 16, whereas the BCS does). Third, and most importantly, the authors had to rely on one-shot measures of parents' and children's earnings. One-shot measures of earnings are known to be quite poor measures of permanent income because they pick up transitory shocks that are uncorrelated with permanent income. Indeed, estimates of intergenerational earnings elasticity vary considerably depending on the number of years over which parents' and children's earnings are averaged; the more years, the higher the elasticity. Erikson and Goldthorpe argue that Blanden et al. observed a decrease in mobility between the two time-periods not because mobility actually decreased but because the transitory component of measured income decreased.

Rather than employing one-shot measures of earnings as indicators of social position, Erikson and Goldthorpe recommend employing social class, which is generally not subject to measurement error due to transitory shocks. In this spirit, they re-analysed the NCDS and the BCS by first assembling mobility tables based around parents' and children's respective social classes, and then fitting a number of log-linear models to them. In particular, they fitted an independence model, which postulates that there is no association between parents' class and children's class in either time period; a constant social fluidity model, which postulates that there is an association between parents' class and children's class but it does not vary between the two time-periods; and a uniform difference model, which postulates that the non-zero association between parents' class and children's class varies between the two time-periods. Erikson and Goldthorpe rejected the independence model, but found that the uniform difference model did not significantly improve upon the fit of the constant social fluidity model. In other words, the association between parents' class and children's class was not higher in the later time period than in the earlier one.

Goldthorpe and Mills, in a paper titled Trends in Intergenerational Class Mobility in Modern Britain: Evidence from National Surveys, 1972-2005, reached the same conclusion as Erikson and Goldthorpe. And so did Goldthorpe and Jackson, in a paper titled Intergenerational Class Mobility in Britain: Political Concerns and Empirical Findings. Moreover, in a working paper titled Has Social Mobility in Britain Declined? New Findings From Cross-cohort Analyses, Bukodi et al. find evidence that social class mobility in Britain has increased over the last few decades, at least for women.

Is it inherently good to have a child?

I’m currently reading Selfish Reasons to Have More Kids by Bryan Caplan. The principal argument of the book is that parents in developed countries can have more children than they are currently having without incurring costs themselves or imposing (substantial) costs on others, and therefore should do so. One of Caplan’s premises is that, excluding any externalities that having a child might produce, it is moral to have a child for the simple reason that you are giving life to someone who otherwise wouldn’t be alive. To illustrate, during the course of his discussion of IVF, Caplan writes, “It is good to exist. The clearest beneficiary of any life-giving technology is the child himself, who would almost certainly be glad to be alive.” In this post, I want to briefly discuss an interesting implication of Caplan’s premise.

In the book, Caplan—based on Julian Simon’s argument from The Ultimate Resource—assumes that, on average, having a child produces an externality with a zero or positive sign. However, he acknowledges there are situations outside the average where having a child produces a negative externality. To take a very simple case, he writes, “According to the Korean adoption study, for example, Mary’s firstborn child will finish six fewer months of education because she gave him four siblings.” Assuming (somewhat unrealistically) that the education externality conferred by siblings is linear, each additional sibling reduces the firstborn’s total education by 1.5 months.

The natural-rights conception of morality asserts that it is immoral to violate someone else’s natural-rights, say by conferring a negative externality on her. Under this conception of morality, regardless of any inherent positive good having a child does, it is immoral to have one if doing so confers a negative externality on at least one person.

In contrast, the utilitarian view of morality asserts that it is immoral to take any action that does more harm than good, produces more pain than pleasure, or yields more preference-dissatisfaction than preference-satisfaction. Under this view of morality, the implications of Caplan’s premise are slightly more interesting. In particular, if having a child confers a net negative externality then whether one should have a child depends on the relative sizes of the net negative externality and the inherent positive good. If the latter is greater than the former, one should have a child; otherwise one should not. Caplan does not specify how much inherent good he thinks giving someone life does, so it is not clear how large the negative externality would need to be to make having a child immoral. In any case, it follows that moral behaviour will generate higher costs for third parties in a moral universe where Caplan’s premise is true than in a moral universe where having a child does not do any inherent positive good.

Incidentally, the book is highly readable and cites lots of interesting research. Finally, David Friedman has an interesting post about externalities from having children here.

Wednesday, 10 April 2013

Should people making of fun of Thatcher's death be punished by the state?

In the last year or so, a number of individuals in the UK have been punished by the state for making offensive jokes on Twitter or Facebook. The blogger Liberty Scott has an excellent post about this trend. Like Liberty Scott, I believe in free speech. Therefore, I don't think the state should punish anyone for making an offensive joke in a public forum. Such behaviour, I would argue, is best regulated through informal social sanctions. If somebody posts something offensive on her Twitter page, others can then condemn her behaviour as vile or hurtful themselves.

Following Margaret Thatcher's death on Monday, a great many people have been posting offensive jokes about her on Twitter and Facebook. For example, here is one of the least rude Thatcher jokes that I came across. There doesn't seem to be any obvious difference between some (but not all) of these jokes and the ones that have landed people with jail sentences (and other punishments) in the recent past. Which brings me to my main point: if the state was consistent, it would punish all those people who have publicly made (sufficiently) offensive jokes about Thatcher. Presumably, one reason it has not done so is that it would simply not be feasible to punish so many people. Indeed, is it really possible to enforce the germane law in a non-arbitrary way? I would claim that it isn't.