Search
Right Hook: The Tactics of Conservative Criticism

Right Hook: The Tactics of Conservative Criticism

Add to Reading List
Add to Reading List

Over the past few years, there have been some big hitting books from the left criticising inequality, capitalism and 'free market' economics or neoliberalism. Naturally, these books have received a lot of criticism from the right. However, sometimes it seems that this criticism is overzealous: an attempt not merely to question the book, but discredit it entirely, and accuse the authors of various misrepresentations of facts and people along the way. Now, while it's true that some arguments or ideas are essentially 'just wrong', and that some proponents of certain ideas can be intellectually dishonest, the frequency with which these accusations are made is alarming, and I believe they are often mistaken.

In this post I'm going to look at three books that I believe have received such treatment: The Spirit Level: Why Equality is Better for Everyone by Kate Pickett & Richard Wilkinson, Chavs: The Demonisation of the Working Class by Owen Jones, and The Shock Doctrine: The Rise of Disaster Capitalism by Naomi Klein.

I'm going to review the books and some of the criticisms they've received in the hope of leveling the debate, and showing that they are far more credible books than the reviews might have you believe.

The Spirit Level

As anyone who's been around in politics for the past few years will know, The Spirit Level (TSL) is a book that caused quite a stir with its claim that inequality is the root cause of many Bad Things (hereafter referred to as 'The Spirit Level Relationship' or TSLR; I will also refer to the authors as 'W & P'). The authors go through the data, finding negative correlations between inequality and numerous social indicators: across time, between countries and between US states. They also discuss the mechanisms through which inequality might cause these problems, typically relating inequality to stress and status. For example, people who have low material and job status suffer from higher stress and as such are more prone to health problems; similarly, young people with a low status are more likely to get involved with gangs and crime as a way of emulating some sort of career progression or finding an identity. Since its release, the book has received heavy criticism from the right - three rebuttals were produced, all of which claimed to have discredited the research. Naturally, W & P have responded to the criticisms, and the debate continues.

Critics generally have much to say on the issue of causality. The idea that W & P simply confuse correlation and causation could not be held by anyone who has read the book properly, so I will not discuss it. However, there are also more substantive criticisms of W & P's purported causal links, such as the charge that W & P are too vague, and switching between inequality, status and stress without really nailing down the causality. For example, Christopher Snowdon, author of The Spirit Level Delusion, believes that W & P may show that stress and low status cause illness, but they have not shown it is inequality which causes this stress. He argues that W & P "fail to demonstrate that inequality is linked to either the mechanism (stress) or the outcome (disease)." He points out that one of their purportedly status-related illnesses, heart disease, is not actually correlated with inequality.

However, while this may be true, heart disease is far from the only mechanism W & P outline, and they do show that both inequality & stress are related to alcohol abuse. They also cite UNICEF children's report, which explores causal links between inequality, stress and health thoroughly for children, and discuss the Whitehall Reports, which do something similar for low paid, low status civil servants. W & P also highlight a couple of quasi-experiments which are consistent with TSLR, such as the changes in inequality and health outcomes post World War 2 in the US and Japan. It's true that W & P sometimes leave the discussion less complete than it should be, but overall, their discussion of causality is far more credible than is sometimes made out.

The criticisms of W & P's regression analysis often focus on the cross country data. However, in my opinion, the strongest points in the original are not related to the overall cross country regressions. First, W & P repeatedly point out that TSLR is strong within similar groups of countries: Scandinavia, the English speaking countries and Spain/Portugal (as well as the PIGS as a whole, if you check) all seem to confirm the relationship. Second, the critics pay little attention to the data from US states, where all but one (crime) of TSLRs hold. Third, W & P have pointed out that the health relationships (which are among the most contentious) have also been shown to hold with multivariate analyses within the provinces of RussiaChileChina and Japan. If I were a critic, I would pay more attention to these areas, where 'all else is equal', as they are far more important pieces of evidence.

In any case, the criticisms of the cross-country data can be overblown. A frequent criticism is of the measures W & P use. They use the 20:20 ratio measure of inequality, which compares the incomes of the rich and poor directly, but critics such as the Tax Payer's Alliance (TPA) claim that if you use different measures of inequality such as the Gini coefficient, TSLR disappears. However, the authors have cross checked their data against both the UN and OECD Gini, as well as the alternative 10:10 ratio, and found that most (though not all, particularly with the OECD Gini) of their findings remain statistically significant for every measure. Furthermore, W & P stick to internationally comparable measures for other data, such as mental illness, where differences in legal tolerance and reporting methods can render comparison difficult.

Inevitably, there are some areas where the critics simply fall into the trap they often accuse W & P of: cherry picking and bad statistics. W & P have argued - and it bears repeating - that while they stick to their chosen methodology throughout TSL, the criticism has a decidedly ad-hoc feel to it, which detracts from its coherence. The methodology of the health section of the TPA report fits this description. First, the data for the measure of inequality they use is unavailable for 9 out of their 26 countries, simply too big a gap for the analysis to be credible. Second, they use inappropriate health measures, such as cancer (which W & P have explicitly argued is not correlated with inequality), and alcohol/tobacco use (which again is not correlated with inequality) rather than abuse (which is). Another seemingly bizarre recurrence is when critics make out the exclusion or inclusion of a particular country is some sort of trick, but fail to note that including it scarcely alters TSLR. This is true of Slovenia in general, Singapore with respect to mental illness, and so forth.

Peter Saunders' criticism in particular was so poor that he received his own rebuttal from an independent author, Hugh Noble, who was basically forced to give Saunders a statistics lesson, focusing on two major points. First, Saunders' reason for removing certain 'outliers' was not only statistically unjustified in the first place; he was completely inconsistent, removing some countries but not removing others which were even larger 'outliers'. Second, Saunders' multivariate regressions used variables that were not 'independent', something that any statistician will know completely undermines the analysis. More generally, Noble notes that even though a great fuss has been made about how the USA inflates W & P's results: "No one...has been able to show that the statistical data about the USA is actually wrong. The USA is a very unequal society and it is afflicted by a plethora of social problems." Lest people think that Noble's approach is one-sided, he goes on to note several problems he has with TSL, many of which I agree with (such as the idea that increasing equality will benefit rich people directly - a losing battle if there ever was one).

Overall, the most authoritative summary of TSL debate was the independent Rowlingson Report. This concluded that "there is a correlation between income inequality and a range of health and social problems", but that causation was of course far harder to establish and further research was needed. Rowlingson also made the interesting point that "some research suggests that inequality is particularly harmful after it reaches a certain threshold", an idea that surely warrants investigation. The report includes a pertinent quote from Chris Jencks, which sums up the debate quite well: "the social consequences of economic inequality are sometimes negative, sometimes neutral but seldom – as far as I can discover – positive." The critics may give reason to doubt some of W & P's methods and conclusions, but they do not alter this overall picture.

Tomorrow

Right Hook: The Tactics of Conservative Criticism (pt.2) / Chavs: The Demonisation of the Working Class by Owen Jones

In Chavs, author Owen Jones argues that media caricatures and opportunistic politicians have helped shape the image of the British working class as feckless, scrounging criminals, or 'chavs'. We direct a level of ire toward this people that would be unthinkable were the word 'chav' replaced with 'black', 'gay' or even 'pleb', which alone shows how unacceptable it is. Yet this vilification of 'chavs', combined with the fact that many of us consider ourselves 'middle class', obscures the simple fact that a majority of people in the UK fit a fairly reasonable definition of working class, and, on top of this, these people are not work shy louts...


JOIN PIERIA TODAY!

Keep up to date with the latest thinking on some of the day's biggest issues and get instant access to our members-only features, such as the News DashboardReading ListBookshelf & Newsletter. It's completely free.

Comments

Please read our Community Guidelines before posting
Unlearning Economics

@luis

Yeah, that’s fine: a silly bit of rhetoric from me. However, Noble outlines why Saunder’s particular choice of variables often undermined his analysis, starting from section 2.20.

@Christopher

“That’s a pretty lame start and it’s rather arrogant to assume that people who have written extensive critiques of TSL are less familiar with the book than yourself.”

The point I was getting at with causality is that simple blanket statements about confusing correlation and causation will not suffice: for example, both Tino Sanandaji and the TPA make the basic point that 'correlation is not causation'. I also saw it in a lot of discussions of the book on twitter etc. I simply do not think this is good enough, as W & P discuss causality extensively: the debate should be had at their purported causal links rather than with blanket assertions about correlation.

“Correlation/causation isn’t the main issue with TSL because most of the time the correlations they claim exist, do not”

Well, you take this as a given, but I haven’t found much of the critic’s analysis convincing based on the points I make above about selectively picking and choosing countries. For example your choice of countries is somewhat puzzling. Portugal may be at the low end of the spectrum for rich countries, but if anything this would justify excluding it rather than including Hungary, which is simply so poor it does not fit the definition of a rich country. Singapore & Hong Kong, in my opinion, should not be included at all, as they are city states (HK isn’t even technically a country), and you even point out that they fare very well on these indicator, contributing substantially to invalidating TSLR. You attribute this to a similar culture to Japan but I’d require more convincing on this point. Czech, S. Korea and Slovenia are fine choices, although there may be a case for excluding former communist countries (analysis within these countries might work, too, though they are perhaps still too poor). W & P are incredibly defensive about their approach and I wish they’d be more flexible, but I am not convinced they are cherry picking: they simply choose their methodology and stick to it; in fact, they even include Singapore when it is surely a candidate ripe for elimination and works against their analysis.

"W & P are not able to present any medical explanation for how inequality causes infant mortality (they don’t even try)"

On p.210 they have a section called "early experience" which explores this. It seems that the causal link is fairly obvious: relatively poorer families are more likely to have worse diets, experience child abuse and also less healthy parents will generally have less healthy children. That UNICEF report looks at this in depth - though you are correct that I made a mistake by linking to the 2013 report, the 2007 version makes some similar points about relative poverty and the irrelevance of per capita GDP for child well being.

“I don’t recall them doing this. Page number?”

Check out p.183 for a discussion of TSLR within ethnically homogeneous groups. I may also follow up on this on my blog at some point with some regressions within the similar groups of countries.

“No they don’t. They don’t even try.”

In their preliminary points in their response to you, they note their study on deaths from alcoholic liver disease among US states and inequality, and also note that alcohol abuse has a strong social gradient. What's more, both alcohol and drug abuse (which I should have included there) are also included on their index, and they have a chapter on mental health and general drug abuse. You mention that their measures on mental health are not comparable, but they get similar results for US states. You also say the data from US states are “noisy” but as they have noted in question 2 of their response to the TPA, many of the purported demographic explanations of health trends in US states have been called into question.

“The US graphs...just show that poorer states tend to do worse than wealthy states.”

This point is well taken. The claim that “equality is better for everyone” is limited at best and at times it’s not clear what W & P are saying. It is worth noting that in section 3 of her report, Rowlingson makes a few tentative comparisons between the US, UK and Sweden and notes that people from each income bracket do better than their counterparts in the more equal country. However, this is only a small sample of evidence, and it is entirely feasible that increasing equality would be ‘better’ overall because it increased the lot of the poor more than it decreased the lot of the rich. This point has been made by quite a few people and I think it is a far more reasonable claim to make about the efficacy of equality.

“If you were a critic, you would assess W & P’s main argument, which focuses much more on countries than on regions”

I appreciate that yes, you would criticise the work as a whole. What I was really suggesting is that I think the more fruitful debate to be had is not in the cross country regressions, as these seem to be incredibly malleable. I was really pondering on how best to move the debate forward.

“If, as you concede, inequality isn’t linked to either of them, there’s simply no way inequality could have enough effect on life expectancy for the difference to be shown in cross-country comparisons.”

Well, as Rowlingson – and also the Whitehall reports – discuss, there are multiple stress-related illnesses that could have a discernible impact. What’s more, mental health is also associated with numerous illnesses and reduced life expectancy, as well as with inequality (as I noted above).

“An independent author, eh? How impressive. The only thing he’s authored is his ‘rebuttal’.”

I don't really think this dismissal of Noble is fair; you should focus on the substance, especially considering that he makes relevant points about the use and abuse of outliers and independent variables, as well as others. Saunder's removal of outliers was one of the worst pieces of statistical analysis in the entire debate: he was not only inconsistent; the rationale for removing them was statistically unjustified in the first place. Noble points this out repeatedly, and W & P have also called him out on it.

"Oh, and I’m not a conservative."

Yeah, I didn’t write the title of this piece.

“The idea that W & P simply confuse correlation and causation could not be held by anyone who has read the book properly, so I will not discuss it.”

That’s a pretty lame start and it’s rather arrogant to assume that people who have written extensive critiques of TSL are less familiar with the book than yourself. W & P’s discussion of causality is brief and superficial. They just say that their findings are unlikely to be the result of chance (which doesn’t address the point) and mention a couple of supporting anecdotes before quickly turning to the question of reverse causation (none of their critics say the problem is reverse causation).

Correlation/causation isn’t the main issue with TSL because most of the time the correlations they claim exist, do not. There are only a few examples of genuine correlations - infant mortality, foreign aid and recycling - and there are indeed big questions of causality there. W & P are not able to present any medical explanation for how inequality causes infant mortality (they don’t even try) and it is fairly obvious that government is the third variable for recycling and foreign aid (ie. governments that prioritise income equality also make recycling compulsory and give more in foreign aid). W & P have never responded to substantive points like this and it doesn’t look like you will either.

“they do show that both inequality & stress are related to alcohol abuse”

No they don’t. They don’t even try.

“They also cite UNICEF children's report, which explores causal links between inequality, stress and health thoroughly for children”

No it doesn’t. There’s no discussion of inequality in the 2007 report W & P cite. You link to a report published several years later which was influenced by TSL.

“W & P repeatedly point out that TSLR is strong within similar groups of countries: Scandinavia, the English speaking countries and Spain/Portugal (as well as the PIGS as a whole, if you check)”

I don’t recall them doing this. Page number?

“the critics pay little attention to the data from US states”

The US graphs are incredibly noisy and just show that poorer states tend to do worse than wealthy states. W & P dismiss the possibility that poverty, not inequality, has an impact on problems in the US early on in the book, but this is little more than assertion. The graph they use to make their point (p. 22 of paperback) actually shows that all the best performing states are rich and all the poor states perform badly. The third variable of low income affects all their US graphs.

“If I were a critic, I would pay more attention to these areas, where 'all else is equal', as they are far more important pieces of evidence.”

If you were a critic, you would assess W & P’s main argument, which focuses much more on countries than on regions. You’d also have to critique the stuff they actually put in the book, not some studies they mentioned later. You link to Chapter 10 of my book in this post so you must know that there are lots of studies saying different things.

Incidentally, in Britain the more equal places tend to be poorer and the less equal places tend to be richer. This is the opposite of the situation in the US and may explain why W & P ignore regional inequality in the UK. The UK’s most equal city is Sunderland.


“They use the 20:20 ratio measure of inequality...”

They use no less than five different measures of inequality at different times and you have to ask why. The obvious one to use is Gini, which is available for everywhere, but for some reason they only use that in the US. You’re right that it doesn’t matter enormously which measures are used although it does matter to them that Japan be at the most equal end (some measures find it in the middle). TSL basically shows that Japan and Scandinavia do better under some carefully selected outcomes than the Anglo-Saxon countries. For this reason, it’s important to them that Japan be very equal - it’s also important to exclude Hong Kong, South Korea and - often - Singapore, because those countries give the game away.

“W & P stick to internationally comparable measures for other data, such as mental illness, where differences in legal tolerance and reporting methods can render comparison difficult.”

It’s now my turn to question whether you’ve read my book, or even their book. Their graph on mental illness welds together several different studies which have different methodologies. Even W & P coyly admit that these studies “are not strictly comparable”. This is an understatement.

“Second, they use inappropriate health measures, such as cancer (which W & P have explicitly argued is not correlated with inequality)”

Cancer and heart disease are by far the biggest causes of death in Western countries. If, as you concede, inequality isn’t linked to either of them, there’s simply no way inequality could have enough effect on life expectancy for the difference to be shown in cross-country comparisons.

“Another seemingly bizarre recurrence is when critics make out the exclusion or inclusion of a particular country is some sort of trick, but fail to note that including it scarcely alters TSLR. This is true of Slovenia in general, Singapore with respect to mental illness, and so forth.”

The point of including countries like Singapore and Slovenia is that they should have been in the analysis in the first place. If you include all the countries that meet W &P’s arbitrary criteria of being richer than Portugal, many of their correlations disappear completely. The addition of a single country is not going to have a massive effect on the graphs (even TSL graphs aren’t that weak), but collectively, this handful of excluded have an enormous impact. You can’t make a serious claim to be rebutting TSL’s critics if you don’t address the fact that most of the findings W & P purport to show rely on cherry-picking countries and cherry-picking problems. You do neither in this post.

“Peter Saunders' criticism in particular was so poor that he received his own rebuttal from an independent author”

An independent author, eh? How impressive. The only thing he’s authored is his ‘rebuttal’. TSL has received ‘rebuttals’ from any number of sociologists, economists, psychologists and other academics, including several people who are sympathetic to their politics. With the exception of Danny Dorling, W & P have received conspicuously little support from academics - many of whom think TSL is an embarrassment to social science - which is why they resorted to publishing essays by random readers on their website.

Oh, and I’m not a conservative.

"Second, Saunders' multivariate regressions used variables that were not 'independent', something that any statistician will know completely undermines the analysis. "

that is at best poorly phrased. To see why, see if you can come up of any examples of any interesting multivariate cross country regressions in which the explanatory variables are independent.

Suppose you dependent variable is Y and your explanatory variables are X1 and X2. I think what Noble means is that if there are potential causal paths X1=>X2=>Y then you have to think carefully about what you results tell you, and experiment with excluding X2, and maybe think about estimating a system of equations, or whatever. You analysis is not "completely undermined" you just have to think about what you are doing.

Twitter Feed

Robert Flood and Miles Kimball on the Status of the Efficient Markets Theory - by @mileskimball http://t.co/cEY5SjDZ1r

Another bid for global expansion suggests M&S has not learned past lessons - http://t.co/ZvOR41sHKU