The Spirit Level Debate - A Summary

January 04, 2011 1:19 PM

In their response to our report about the empirical claims contained in The Spirit Level, the authors have distorted the evidence. Once again, their claims do not stand up to scrutiny, for that reason this article provides a comprehensive response of our own to their response.

One particularly stark example of the problems with Wilkinson and Pickett's use of evidence comes when they cite the work of Nobel laureate James Heckman. The reader is given the impression that Heckman's work contains evidence supporting Wilkinson and Pickett.

Heckman's paper contains nothing of the kind. His study investigates the effect of different parental investments in children on cognitive and non-cognitive abilities, and therefore looks at the effects of different absolute levels of income rather than the effect of income inequality across a society itself. There is thus zero evidence in this paper (or anywhere else in professor Heckman's seminal work) supporting Wilkinson and Pickett's claim that other people having a higher income is bad for you. We contacted Heckman and he said that "[t]his is a misrepresentation of my work".

Wilkinson and Pickett are simply trying to deceive the general public by giving the impression that their claims represent a scientific consensus, whereas their views actually differ from the mainstream understanding in the academic literature.

In an exchange of letters with us in the Wall Street Journal, Wilkinson and Pickett have already backed away from their main claim in the book; that inequality kills. The reason is that both our work and Christopher Snowdon's book The Spirit Level Delusion have clearly shown that the correlation between life expectancy and inequality does not exist. Wilkinson and Pickett can only suggest that it exists by cherry-picking a specific year, definition of inequality and selection of countries.

Now Wilkinson and Pickett are claiming that they are not even responsible for that central claim in their book. They write: "The claim that 'inequality kills' has been made for us, in publishers' publicity and in translation."

In our response we have found many, many examples of errors and misleading claims that call into question the reliability of Wilkinson and Pickett's work.

Another example is their use of low quality data to assess patent intensity. In our initial report, we found that the data regarding a link between patents and inequality in The Spirit Level was inaccurate. Wilkinson and Pickett defend their book, explaining that the data is from the World Intellectual Property Organization. But they actually give a link from www.nationmaster.com, where the data is clearly wrong.

Wilkinson and Pickett explain to us that patents per capita are roughly the same for the US and for Portugal. Anyone who goes to the actual source - to actually look up data from the World Intellectual Property Organization rather than just log on to www.nationmaster.com - would find that the US in fact has close to 50 times as many patents per capita as Portugal. It is telling that Wilkinson and Pickett did not do so when confronted with the finding that the US is no more innovative than Portugal, which is absurd at face value. Then again, these are the same authors who believe Cubans have a longer lifespan than Americans.

We hope that those who are interested in the ongoing debate relating to The Spirit Level will read this document, to understand why it is so wrong to trust Wilkinson and Pickett’s findings and assume that they are basing their arguments on a broad understanding of an uncontested scientific literature.

Our Original Questions

Q1. You claim to present an overview of the research on health and inequality, yet leave out the scientifically most heavyweight survey of the field, Princeton Professor Angus Deaton's article in the prestigious Journal of Economic Literature. Is this simply because Deaton finds no robust relationship between life expectancy and income inequality among the rich countries? (Deaton, A. S. 'Health, inequality and economic development', Journal of Economic Literature, May 2001)

Q2. You base much of your thesis on the relationship between inequality and life expectancy within U.S states. Why do you neglect to tell your audience that researchers have found that this relationship vanishes once they control for demographic differences? (Deaton, A.S., D. Lubotsky. "Mortality, inequality and race in American cities and states", Social Science & Medicine, March 2003)

Response from Wilkinson and Pickett

Angus Deaton's 2001 study is far from being the most up-to-date review of inequality and health and the social determinants of health has never been his main field. Look also at more recent work, perhaps particularly from the group at the Harvard School of Public Health, many of which are co-authored by I Kawachi, Professor of Social Epidemiology and Chair of the Department of Society, Human Development, and Health at Harvard (see for instance reviews such as Kondo N, Sembajwe G, Kawachi I, van Dam RM, Subramanian SV, Yamagata Z. Income inequality, mortality, and self rated health: a meta-analysis of multilevel studies. British Medical Journal 2009:339).

In 2006, we published a much more comprehensive review than Deaton's, taking into account the five years of research published since his 2001 paper, and based on close to 200 studies. Other 'heavyweight' economists, including Nobel laureates, have also written about the significance of inequality for wellbeing and human capital formation. Since Deaton's paper several of the issues he raised have been the subject of further research, including the idea that the relationship between inequality and health in the 50 states is actually attributable to the proportion of each state's population which is African-American. The same point was once raised in relation to the pattern of violence among the states. Both have now been shown to be inaccurate. (See 38-41. We hope the TaxPayers' Alliance will inform their readers of these more recent findings to avoid further misunderstanding.)

The reason why the initial confusion over the role of ethnicity arose is because in those states where there is a larger African American population there is also a bigger income gap between blacks and whites. But in the states with a higher proportion of African Americans it is not only black health which is worse: white health is also poorer. The issue is not of course that skin colour determines health. Instead, ethnicity matters when skin colour becomes a marker of social status which attracts the same discrimination and downward prejudice that low social status and deprivation have always attracted. The evidence on violence is similar. Relations between inequality and violence exist in both Southern and Northern states. Rates of homicide perpetrated by white men are related to income inequality even when inequality is measured only among whites. To suggest removing or controlling for the proportion of the population which is African American in each state is analogous to saying one should look at the effects of inequality only after taking out the disadvantaged.

We do not of course base our thesis on a single relationship between inequality and life expectancy within US states. Given that here are now over 200 studies testing this relationship there is no possible reason for doing so.

Our Response

The reference to "'heavyweight' economists, including Nobel laureates", refers to the work of Nobel Prize winner James Heckman (that source is given in the document on their website). The reader is given the impression that Heckman's work contains evidence supporting Wilkinson and Pickett.

However Heckman's paper contains nothing of the kind. His study investigates the effect of parental investment on cognitive and non-cognitive abilities, and how differences in those investments drive inequality, not the effect of inequality. There is zero evidence in this paper (or anywhere else in professor Heckman's seminal work) supporting Wilkinson and Pickets claim that other people having higher incomes is bad for you. When we contacted Heckman, he said that "[t]his is a misrepresentation of my work".

Note Wilkinson and Pickett's choice of words. They write that Heckman has "written" about inequality and health, which is of course technically true. But what they don't tell the readers is that while he has indeed written about these issues, he has not found any evidence supporting the claims they have made.

The most robust article about income inequality and health was written by Angus Deaton, and concluded that no causal relationship has been established between nequality and life expectancy. Yet in their book popularising research into this issue for the general public, Wilkinson and Pickett never cited or acknowledged the existence of this study.

When pushed, Wilkinson and Pickett dismiss Deaton as not competent to pass judgement on the "social determinants of health". But Deaton's article was published in the Journal of Economic Literature (JEL), the most authoritative journal in economics for review articles. One of the purposes of this journal is to invite scholars that rest of the field consider one of the prime experts in that field to summarise the views of the profession.

Wilkinson and Pickett may not consider Deaton enough of an expert to discuss the relationship between income inequality and health, but the economic profession obviously did, in that Deaton was invited to write the overview article about "Health, inequality, and economic development" in the JEL.

Deaton's article currently has more than 5 times as many impact-factor weighted citations as Wilkinson and Pickett's own review article. Wilkinson and Pickett may not consider Deaton an expert, but the academic community clearly does.

Wilkinson and Pickett then claim that Deaton's study (published in 2003) should be ignored because it is not "up-to-date", (as opposed to Wilkinson and Pickett's own study, which was published in 2006).

This reveals the second biggest problem with Wilkinson and Pickett’s work, their failure to understand stringent methodological standards. Deaton's conclusion that health and income distribution were not related was based on methodological weaknesses with studies that claimed it did, and the inability to establish any causative link between inequality and life expectancy.

One symptom of this problem is that the correlation between health and inequality that is sometimes found vanishes in more careful studies that control for background variables (and in any case, since not all background variables can be measured and controlled for, there is a need for experimental methods to convincingly link inequality and poor health).

Modern health economists simply have higher criteria than Wilkinson and Pickett, and require more than simple correlations. This is why the consensus view of the literature is the opposite of that reported by Wilkinson and Pickett.

That methodological problem was not addressed in the few years that passed between Deaton's study and Wilkinson and Pickett's study. No study using causality criteria in mainstream economics causally links inequality to poor health, whether published before or after 2003. The claim that Deaton's study is outdated is simply false.

Suppose for a moment that the Wilkinson and Pickett study is right and Deaton's wrong. That should mean it would be easy for the authors to refute Deaton. Instead, Wilkinson and Pickett simply ignore the most heavyweight review study in academia and do not alert their readers that their finding is contested.

There are very recent studies which do not agree with Wilkinson and Pickett’s conclusions. For example in the paper "Inequality and mortality: Long-run evidence from a panel of countries" (A Leight and C Jencks, Journal of Health Economics, 2007) it is noted that "the relationship between income inequality and health is either non-existent or too fragile to show up in a robustly estimated panel specification."

The studies by Deaton and Leight and Jencks have a full grasp of the dangers of using correlative evidence to assess causality. Thus to the extent that any consensus exists within the discipline of health economics about the relationship between income distribution and life expectancy, it is that no such causal effect has yet been established. Yet Wilkinson and Pickett lend these important articles zero weight, neglecting to even cite them, despite their authoritative position in the literature.

It is interesting that Wilkinson and Pickett refer to a paper by Kondo et al. in their response, giving the impression that the research in question supports their view that inequality is at least partly responsible for most, if not all, social ills. Contrary to Wilkinson and Pickett's insinuation, the Kondo et al. review never claims that they have causally linked inequality with poor health, instead consistently referring to an "association".

They further note that the findings need to be interpreted with "caution given the heterogeneity between studies, as well as the attenuation of the risk estimates in analyses that attempted to control for the unmeasured characteristics of areas with high levels of income inequality."

(Income inequality, mortality, and self rated health: a meta-analysis of multilevel studies, British Medical Journal, 2009)

In layman's terms, once background variables are controlled for, the link between inequality and health can vanish, exactly as was found in the previous research. Wilkinson and Pickett persistently confound causation and correlation.

Wilkinson and Pickett also refer to research by Robinson et al. (Inequality, race, and mortality in U.S. cities: a political and econometric review of Deaton and Lubotsky, Soc Sci Med, 2009) in arguing that Deaton and Lubotsky are wrong in claiming that the relationship between mortality and income inequality vanishes once demographics are controlled for. They don’t find room to mention that Deaton and Lubotsky have in turn responded to that article!

In the response from Deaton and Lubotsky they say:
"Deaton and Lubotsky (2003) found that the robust positive relationship across American cities between mortality and income inequality became small, insignificant, and/or non-robust once they controlled for the fraction of each city's population that is black. Ash and Robinson (Ash, M., & Robinson D. Inequality, race, and mortality in US cities: a political and econometric review. Social Science and Medicine, 2009) consider alternative weighting schemes and show that in one of our specifications, in one data period, and with one of their alternative weighting schemes, income inequality is estimated to be a risk factor. All of our other specifications, as well as their own preferred specification, replicate our original result, which is supported by the weight of the evidence. Conditional on fraction black, there is no evidence for an effect of income inequality on mortality."

(Income inequality and mortality in U.S. cities: Weighing the evidence. A response to Ash, Soc Sci Med, 2009)

Wilkinson and Pickett only refer to research that supports their own ideas, completely neglecting other work regardless of how important it is to a proper understanding of the issue.

Those interested in the issue should note that in 2002 professor of public health Johan Mackenbach wrote in a survey:
"Overall these papers reinforce the idea that the evidence for a correlation between income inequality and the health of the population is slowly dissipating. There is very little confirmation of such a relation outside the United States. Within the United States it has still to be convincingly demonstrated that it is not due to curvilinear individual level relationships and confounding."

(Income inequality and population health, British Medical Journal, 2002)

The general public - the target audience for The Spirit Level - cannot be expected to be aware of the state of research in the field. Wilkinson and Pickett exploit the trust of their readers and give them the impression that their claims represent consensus science, when the opposite is closer to the truth.

We also note that Wilkinson and Pickett write that "there are now over 200 studies testing [the relationship between inequality and life expectancy]." They give a similar response to another one of our questions, referring to their 2006 review noted above. But in the review they themselves write:

"We identified 168 analyses in 155 papers reporting research findings on the association between income distribution and population health"
(Income inequality and population health: a review and explanation of the evidence, Social Science & Medicine, 2006)

It is unclear how this fits with Wilkinson and Pickett’s claim of over 200 studies dealing with life expectancy and equality.

Our Original Question

Q3. Correlation is not causation. This is true both for simple relationships and with multiple variables. Do you have any studies that actually establish a relationship between life expectancy and inequality, based on exogenous variation of inequality, quasi-experiments or any other well identified source of variation?

Response from Wilkinson and Pickett

Indeed, correlation is not necessarily proof of causation. However, as epidemiologists, we are trained in researching causal relationships within an observational framework. One of our critics has admitted that "sociology has been remarkably inept at providing us with the evidence and tools to create a better society". We agree, but epidemiology has been much more successful in uncovering causal influences on population health and in pointing the way to effective public health policy. Epidemiologists have been able, within an observational framework, to show that: smoking causes lung cancer; sleep position affects babies' risk of dying; social status and social networks have a profound impact of on people's risk of chronic disease, etc...

We discuss the epidemiological criteria for the establishment of causality in our book. One example might strike readers as a useful illustration. At the end of the Second World War, the USA was a very equal country, and ranked high on population health; Japan was a very unequal country and ranked very poorly on population health. Since then, these two countries have switched their relative positions: the USA is now very unequal and ranks very low on health; Japan became much more equal and its life expectancy increased faster than any other developed country till it had the highest life expectancy in the world. See also the study by Clarkwest et al referenced in point 4 below 18. There are a number of papers dealing with changes over time, with path analysis, and with causal ordering. In addition, a number of the associations found in observational studies of humans, including biological measurements, have been supported by genuine experiments on non-human primates where social status can be manipulated while material standards are kept constant.

Our Response

Wilkinson and Pickett avoid answering our question. They only cite one article, which relates to China. (Provincial income inequality and self-reported health status in China during 1991-7, J Epidemiol Community Health, 2006).

It is surprising that this paper is considered relevant by Wilkinson and Pickett when it only has 6 citations according to Google Scholar whilst Deaton's paper, with almost 100 times that many citations, is not. Also, why do Wilkinson and Pickett suddenly turn to a study focused on China when their analysis (particularly in The Spirit Level) is focused on OECD countries?

Wilkinson and Pickett do write that there "are a number of papers" which show a causal link. Why do they not refer to these papers? We repeat the question: Do you have any studies that actually establish a relationship between life expectancy and inequality, based on exogenous variation of inequality, quasi-experiments or any other well identified method of studying variation?

Wilkinson and Pickett rely on simple correlation studies. But, if we observed a correlation between headaches and taking aspirin, we would not conclude that aspirin causes headaches. Similarly, that income distribution is occasionally found to be correlated with various social ills should not come as a surprise.

It is, in fact, difficult to imagine a social problem that does not plausibly depress the income of the impacted group in some way, thereby increasing income inequality and giving rise to a spurious correlation. Repeating the flawed methodology on numerous occasions will not magically turn correlation into causation.

In the modern social sciences, researchers tend to understand the need to conduct careful statistical analysis when dealing with complex economic phenomenon such as income inequality. Wilkinson and Pickett do not rely on statistical methods that actually separate simple correlation and causality, even though the question they are engaging in is particularly prone to complex ties between variables.

Our Original Questions

4. Your most famous claim is that "inequality kills". Yet, using OECD life expectancy data, UN life expectancy data, OECD Gini and UN Gini, with different selections of countries, in several specifications, we again and again fail to replicate your result and find any statistically significant relationship between life expectancy and inequality. Is the explanation that you have relied on cherry picking – using the exact selection of measures, countries and year where such a correlation can be shown to exist?

5. Your initial defence for the lack of a statistically significant relationship between life expectancy and inequality from OECD data was that we should look at the working age population. Do you have any further defence, given that the OECD data shows no statistically significant relationship for the population between the ages of 15 and 60?

6. If inequality (rather than poverty) is strongly related to poor health, why can we not find any statistically significant relationship between inequality and health outcomes, as measured by the OECD for 16 of 19 health variables?

Response from Wilkinson and Pickett

The claim that "inequality kills" has been made for us, in publishers' publicity and in translation. However the weight of the evidence from studies of either infant or adult mortality, among both rich and poor countries, the American states, the regions of Russia, the provinces of China, the counties of Chile and many more suggests it does. Relationships between income inequality and life expectancy have been repeatedly demonstrated since 1979 13. There are over 200 tests of this link, internationally and in the US states, and the vast majority of studies confirm the adverse impact of inequality on health.

However, after periods in which income distribution has changed rapidly, the cross-sectional international association between income inequality and life expectancy have sometimes seemed to disappear 15 – only to reappear later 16. This seems to be because there are substantial lag periods between changes in income distribution and changes in population health. Every cause of death, and death rates in every age group have different lag periods. Death rates in later life are known to be powerfully influenced by experience in early life. Taking this into account it is surprising that relationships between health and inequality have been demonstrated so many times in so many different contexts. But we accept that the inequality/health relationship is one of the weaker associations demonstrated in The Spirit Level – no doubt partly for the reasons just described. Because lag periods are much shorter for infant mortality these relationships have been more consistent as have the association among US states. However, two new pieces of evidence leave little room for doubt as to the veracity of these relationships.

One is a study published in the British Medical Journal – a meta-analysis of multi-level studies of income inequality and health. This shows unequivocally that, even after controlling for individual income or education, inequality is related to significantly higher mortality rates. The second is a study showing that US states with bigger increases in inequality between 1970 and 2000 had less improvement in life expectancy than those with smaller increases.

Our Response

It is telling that Wilkinson and Pickett are suddenly backing away from the prominent claim in their book that inequality kills. We are glad they admit that "inequality kills" is misleading, but it is interesting to note that they put the responsibility on their publishers and the translators. They should have done more to make it clear that claim does not stand up.

After all, an earlier book authored by Wilkinson was marketed with the same byline "Inequality Kills" (see the Guardian review of the book "The Impact of Inequality" in 2005).

And, as we wrote in our fourth question:
"Your most famous claim is that "inequality kills". Yet using OECD life expectancy data, UN life expectancy data, OECD Gini and UN Gini, with different selections of countries, in several specifications, we again and again fail to replicate your result and find any statistically significant relationship between life expectancy and inequality. Is the explanation that you have relied on cherry picking – using the exact selection of measures, countries and year where such a correlation can be shown to exist?"

Wilkinson and Pickett still need to explain why the relationship they show in their book only appears if a specific definition of equality, selection of countries and year is chosen.

They again claim that "over 200" studies establish a link between life expectancy and inequality, which is hard to square with their review as we mentioned earlier.

There is no response to our fifth question. Why did they tell us to look at the working age population, when no correlation exists for this group either? Perhaps most importantly, Wilkinson and Pickett do not respond to our sixth question either. Let us repeat the question: "If inequality (rather than poverty) is strongly related to poor health, why can we not find any statistically significant relationship between inequality and health outcomes as measured by the OECD for 16 of 19 health variables?"

Our Original Question

Q7. If inequality is strongly related to life expectancy, why have the countries with the highest increase in inequality witnessed, on average, higher increases in life expectancy in the last two decades according to OECD data?

Response from Wilkinson and Pickett

Researching changes in inequality and changes in outcomes is difficult, and needs careful thought about lag times. The widening of income distribution which started in the late 1970s or early 1980s; followed the spread of neo-liberal economic and political thinking from the English speaking countries to other countries. While income differences rose rapidly in the 1980s in several English speaking countries, it remained stable in a number of other developed countries before spreading to them a decade or two later. The question is which period are current improvements in health related really to – current increases in inequality or the earlier stability? So you may have the relationships exactly the wrong way round. But if you have good evidence it should be presented in a peer reviewed journal. Clarkwest and colleagues 18 have shown that states with greater changes in inequality between 1970 and 2000 have had less improvement in life expectancy than those with smaller increases.

Our Response

It is of course a very valid point that correlation is not the same as causation. But Wilkinson and Pickett base all of their analysis on correlations (many of which do not stand up to scrutiny). Their logic in this case has implications for their wider work.

Wilkinson and Pickett refer to the paper "Neo-materialist theory and the temporal relationship between income inequality and longevity change" (Social Science & Medicine, 2008). The paper has three citations according to Google Scholar so again it is difficult to see why they refer to it while ignoring the work done by Deaton (only 565 citations).

More importantly, the abstract of the paper Wilkinson and Pickett refer to says:

"Contrary to findings from prior research, analyses also reveal a strong negative association between change in inequality and change in longevity once initial levels of inequality and other state characteristics are controlled."

The paper supports the Wilkinson and Pickett theory, but in its own abstract notes that its findings disagree with the results found in other research. Unsurprisingly, Wilkinson and Pickett only refer to this paper, not those that find different results.

Our Original Question

Q8. Why do you claim that more unequal nations have less creativity, (and that Portugal is as creative as the United States), when data from the World Intellectual Patent Organization shows the opposite?

Response from Wilkinson and Pickett

We used data from the World Intellectual Patent Organisation.

This shows patents per capita for Portugal at 0.6 and the USA at 1.0. In contrast, patents per capita for Japan are 7.8 and for Sweden 30.1.

Our Response

Wilkinson and Pickett describe their source as the World Intellectual Property Organization but the link is to an amateur internet site – Nationmaster. That website gives numbers for patents granted per million people.

The preferred measure for patent intensity given by the World intellectual Property Organization is patent filings per million population, and is available directly from their website. It shows that between 1995 and 2007, the number of patent fillings per million inhabitants in Portugal rose from 8.08 to 23.57. In the US, the corresponding figures are 465.54 and 800.17 for the same years.

Looking instead at patents granted, in 2008 the Portugal total is 241 and the United States of America total is 147,154. Adjusted for population, that is about 470 grants per million people in the United States against around 23 per person in Portugal.

So the latest data – from the World Intellectual Property Organization – shows there were far more patents per person in the US as in Portugal whether measured by patents granted or filed. It appears that Wilkinson and Pickett have obtained the results they did by failing to check back to the original source.

To look up the data directly from the World Intellectual Property Organizations:

1. Go to the site http://www.wipo.int/ipstats/en/statistics/patents/
2. Click on the link "Resident patent filings per million population (1995-2007)" or "Patent grants by country of origin and by office (1995-2008)".

A researcher who finds that the US has an equal patent intensity to Portugal should realise that she or he could be using a flawed data source. They should then at least check their results against the original source (the World Intellectual Property Organization in this case) and not just a site on the web.

Our Original Question

Q9. Why do you claim that more unequal nations have more mental illnesses (and that the United Kingdom has 250% the mental illness level of Germany) when data from the World Health Organisation shows the opposite?

Response from Wilkinson and Pickett

We use W.H.O. data designed to provide comparable estimates of the prevalence of mental illness. Rather than reflecting the use of medical and psychiatric services as the data on which this question is based does, the data we used is based on the scientific collection of data using standardised diagnostic interviews administered to random samples of the population. The question reflects a confusion epidemiologists are taught to avoid at the beginning of their training. The only reason W.H.O. went to the expense of collecting the data we used is that help-seeking behaviour is a very unreliable guide to the prevalence of health problems.

Our Response

After finding such a striking difference in results between the United Kingdom and Germany, it is unfortunate that Wilkinson and Pickett did not note that on another measure, even if it is dependent on help seeking behaviour, the findings are quite different.

With such a small sample of countries the reliability of the correlation derived from the survey is clearly open to question and the numbers treated, within mostly high income modern societies, is far from irrelevant and suggests the opposite conclusion.

Our Original Question

Q10. Why do you claim that "[i]n Sweden, people don't bother to check your tickets on the train or bus" when this is obviously not the case? The American audience reading the Boston Globe might believe you, but anybody who has lived in or visited Sweden will immediately see through the deception.

Response from Wilkinson and Pickett

We have, happily, visited both Sweden and the USA several times recently so our claims are based on our own experiences in those countries. But our experiences are borne out by the evidence on trust.

Our Response

All three authors of this response live in Sweden and have done so for most of their lives. We have used Swedish trains, buses and metros for transportation thousands of times. When you go on a Swedish bus, you show your ticket or buy one. Occasionally, several inspectors board the bus and check that nobody has somehow cheated (by for example buying a ticket to stop A but instead travelling further to stop B).

When you travel with a Swedish train, your ticket will be checked. In the Stockholm metro system, significant investments have been made to stop people from jumping over the barriers without paying for a ticket. And, yet again, random checks are performed on the metro.

We are unsure how the authors of The Spirit Level have come to this conclusion.In their response to our report about the empirical claims contained in The Spirit Level, the authors have distorted the evidence. Once again, their claims do not stand up to scrutiny, for that reason this article provides a comprehensive response of our own to their response.

One particularly stark example of the problems with Wilkinson and Pickett's use of evidence comes when they cite the work of Nobel laureate James Heckman. The reader is given the impression that Heckman's work contains evidence supporting Wilkinson and Pickett.

Heckman's paper contains nothing of the kind. His study investigates the effect of different parental investments in children on cognitive and non-cognitive abilities, and therefore looks at the effects of different absolute levels of income rather than the effect of income inequality across a society itself. There is thus zero evidence in this paper (or anywhere else in professor Heckman's seminal work) supporting Wilkinson and Pickett's claim that other people having a higher income is bad for you. We contacted Heckman and he said that "[t]his is a misrepresentation of my work".

Wilkinson and Pickett are simply trying to deceive the general public by giving the impression that their claims represent a scientific consensus, whereas their views actually differ from the mainstream understanding in the academic literature.

In an exchange of letters with us in the Wall Street Journal, Wilkinson and Pickett have already backed away from their main claim in the book; that inequality kills. The reason is that both our work and Christopher Snowdon's book The Spirit Level Delusion have clearly shown that the correlation between life expectancy and inequality does not exist. Wilkinson and Pickett can only suggest that it exists by cherry-picking a specific year, definition of inequality and selection of countries.

Now Wilkinson and Pickett are claiming that they are not even responsible for that central claim in their book. They write: "The claim that 'inequality kills' has been made for us, in publishers' publicity and in translation."

In our response we have found many, many examples of errors and misleading claims that call into question the reliability of Wilkinson and Pickett's work.

Another example is their use of low quality data to assess patent intensity. In our initial report, we found that the data regarding a link between patents and inequality in The Spirit Level was inaccurate. Wilkinson and Pickett defend their book, explaining that the data is from the World Intellectual Property Organization. But they actually give a link from www.nationmaster.com, where the data is clearly wrong.

Wilkinson and Pickett explain to us that patents per capita are roughly the same for the US and for Portugal. Anyone who goes to the actual source - to actually look up data from the World Intellectual Property Organization rather than just log on to www.nationmaster.com - would find that the US in fact has close to 50 times as many patents per capita as Portugal. It is telling that Wilkinson and Pickett did not do so when confronted with the finding that the US is no more innovative than Portugal, which is absurd at face value. Then again, these are the same authors who believe Cubans have a longer lifespan than Americans.

We hope that those who are interested in the ongoing debate relating to The Spirit Level will read this document, to understand why it is so wrong to trust Wilkinson and Pickett’s findings and assume that they are basing their arguments on a broad understanding of an uncontested scientific literature.

Our Original Questions

Q1. You claim to present an overview of the research on health and inequality, yet leave out the scientifically most heavyweight survey of the field, Princeton Professor Angus Deaton's article in the prestigious Journal of Economic Literature. Is this simply because Deaton finds no robust relationship between life expectancy and income inequality among the rich countries? (Deaton, A. S. 'Health, inequality and economic development', Journal of Economic Literature, May 2001)

Q2. You base much of your thesis on the relationship between inequality and life expectancy within U.S states. Why do you neglect to tell your audience that researchers have found that this relationship vanishes once they control for demographic differences? (Deaton, A.S., D. Lubotsky. "Mortality, inequality and race in American cities and states", Social Science & Medicine, March 2003)

Response from Wilkinson and Pickett

Angus Deaton's 2001 study is far from being the most up-to-date review of inequality and health and the social determinants of health has never been his main field. Look also at more recent work, perhaps particularly from the group at the Harvard School of Public Health, many of which are co-authored by I Kawachi, Professor of Social Epidemiology and Chair of the Department of Society, Human Development, and Health at Harvard (see for instance reviews such as Kondo N, Sembajwe G, Kawachi I, van Dam RM, Subramanian SV, Yamagata Z. Income inequality, mortality, and self rated health: a meta-analysis of multilevel studies. British Medical Journal 2009:339).

In 2006, we published a much more comprehensive review than Deaton's, taking into account the five years of research published since his 2001 paper, and based on close to 200 studies. Other 'heavyweight' economists, including Nobel laureates, have also written about the significance of inequality for wellbeing and human capital formation. Since Deaton's paper several of the issues he raised have been the subject of further research, including the idea that the relationship between inequality and health in the 50 states is actually attributable to the proportion of each state's population which is African-American. The same point was once raised in relation to the pattern of violence among the states. Both have now been shown to be inaccurate. (See 38-41. We hope the TaxPayers' Alliance will inform their readers of these more recent findings to avoid further misunderstanding.)

The reason why the initial confusion over the role of ethnicity arose is because in those states where there is a larger African American population there is also a bigger income gap between blacks and whites. But in the states with a higher proportion of African Americans it is not only black health which is worse: white health is also poorer. The issue is not of course that skin colour determines health. Instead, ethnicity matters when skin colour becomes a marker of social status which attracts the same discrimination and downward prejudice that low social status and deprivation have always attracted. The evidence on violence is similar. Relations between inequality and violence exist in both Southern and Northern states. Rates of homicide perpetrated by white men are related to income inequality even when inequality is measured only among whites. To suggest removing or controlling for the proportion of the population which is African American in each state is analogous to saying one should look at the effects of inequality only after taking out the disadvantaged.

We do not of course base our thesis on a single relationship between inequality and life expectancy within US states. Given that here are now over 200 studies testing this relationship there is no possible reason for doing so.

Our Response

The reference to "'heavyweight' economists, including Nobel laureates", refers to the work of Nobel Prize winner James Heckman (that source is given in the document on their website). The reader is given the impression that Heckman's work contains evidence supporting Wilkinson and Pickett.

However Heckman's paper contains nothing of the kind. His study investigates the effect of parental investment on cognitive and non-cognitive abilities, and how differences in those investments drive inequality, not the effect of inequality. There is zero evidence in this paper (or anywhere else in professor Heckman's seminal work) supporting Wilkinson and Pickets claim that other people having higher incomes is bad for you. When we contacted Heckman, he said that "[t]his is a misrepresentation of my work".

Note Wilkinson and Pickett's choice of words. They write that Heckman has "written" about inequality and health, which is of course technically true. But what they don't tell the readers is that while he has indeed written about these issues, he has not found any evidence supporting the claims they have made.

The most robust article about income inequality and health was written by Angus Deaton, and concluded that no causal relationship has been established between nequality and life expectancy. Yet in their book popularising research into this issue for the general public, Wilkinson and Pickett never cited or acknowledged the existence of this study.

When pushed, Wilkinson and Pickett dismiss Deaton as not competent to pass judgement on the "social determinants of health". But Deaton's article was published in the Journal of Economic Literature (JEL), the most authoritative journal in economics for review articles. One of the purposes of this journal is to invite scholars that rest of the field consider one of the prime experts in that field to summarise the views of the profession.

Wilkinson and Pickett may not consider Deaton enough of an expert to discuss the relationship between income inequality and health, but the economic profession obviously did, in that Deaton was invited to write the overview article about "Health, inequality, and economic development" in the JEL.

Deaton's article currently has more than 5 times as many impact-factor weighted citations as Wilkinson and Pickett's own review article. Wilkinson and Pickett may not consider Deaton an expert, but the academic community clearly does.

Wilkinson and Pickett then claim that Deaton's study (published in 2003) should be ignored because it is not "up-to-date", (as opposed to Wilkinson and Pickett's own study, which was published in 2006).

This reveals the second biggest problem with Wilkinson and Pickett’s work, their failure to understand stringent methodological standards. Deaton's conclusion that health and income distribution were not related was based on methodological weaknesses with studies that claimed it did, and the inability to establish any causative link between inequality and life expectancy.

One symptom of this problem is that the correlation between health and inequality that is sometimes found vanishes in more careful studies that control for background variables (and in any case, since not all background variables can be measured and controlled for, there is a need for experimental methods to convincingly link inequality and poor health).

Modern health economists simply have higher criteria than Wilkinson and Pickett, and require more than simple correlations. This is why the consensus view of the literature is the opposite of that reported by Wilkinson and Pickett.

That methodological problem was not addressed in the few years that passed between Deaton's study and Wilkinson and Pickett's study. No study using causality criteria in mainstream economics causally links inequality to poor health, whether published before or after 2003. The claim that Deaton's study is outdated is simply false.

Suppose for a moment that the Wilkinson and Pickett study is right and Deaton's wrong. That should mean it would be easy for the authors to refute Deaton. Instead, Wilkinson and Pickett simply ignore the most heavyweight review study in academia and do not alert their readers that their finding is contested.

There are very recent studies which do not agree with Wilkinson and Pickett’s conclusions. For example in the paper "Inequality and mortality: Long-run evidence from a panel of countries" (A Leight and C Jencks, Journal of Health Economics, 2007) it is noted that "the relationship between income inequality and health is either non-existent or too fragile to show up in a robustly estimated panel specification."

The studies by Deaton and Leight and Jencks have a full grasp of the dangers of using correlative evidence to assess causality. Thus to the extent that any consensus exists within the discipline of health economics about the relationship between income distribution and life expectancy, it is that no such causal effect has yet been established. Yet Wilkinson and Pickett lend these important articles zero weight, neglecting to even cite them, despite their authoritative position in the literature.

It is interesting that Wilkinson and Pickett refer to a paper by Kondo et al. in their response, giving the impression that the research in question supports their view that inequality is at least partly responsible for most, if not all, social ills. Contrary to Wilkinson and Pickett's insinuation, the Kondo et al. review never claims that they have causally linked inequality with poor health, instead consistently referring to an "association".

They further note that the findings need to be interpreted with "caution given the heterogeneity between studies, as well as the attenuation of the risk estimates in analyses that attempted to control for the unmeasured characteristics of areas with high levels of income inequality."

(Income inequality, mortality, and self rated health: a meta-analysis of multilevel studies, British Medical Journal, 2009)

In layman's terms, once background variables are controlled for, the link between inequality and health can vanish, exactly as was found in the previous research. Wilkinson and Pickett persistently confound causation and correlation.

Wilkinson and Pickett also refer to research by Robinson et al. (Inequality, race, and mortality in U.S. cities: a political and econometric review of Deaton and Lubotsky, Soc Sci Med, 2009) in arguing that Deaton and Lubotsky are wrong in claiming that the relationship between mortality and income inequality vanishes once demographics are controlled for. They don’t find room to mention that Deaton and Lubotsky have in turn responded to that article!

In the response from Deaton and Lubotsky they say:
"Deaton and Lubotsky (2003) found that the robust positive relationship across American cities between mortality and income inequality became small, insignificant, and/or non-robust once they controlled for the fraction of each city's population that is black. Ash and Robinson (Ash, M., & Robinson D. Inequality, race, and mortality in US cities: a political and econometric review. Social Science and Medicine, 2009) consider alternative weighting schemes and show that in one of our specifications, in one data period, and with one of their alternative weighting schemes, income inequality is estimated to be a risk factor. All of our other specifications, as well as their own preferred specification, replicate our original result, which is supported by the weight of the evidence. Conditional on fraction black, there is no evidence for an effect of income inequality on mortality."

(Income inequality and mortality in U.S. cities: Weighing the evidence. A response to Ash, Soc Sci Med, 2009)

Wilkinson and Pickett only refer to research that supports their own ideas, completely neglecting other work regardless of how important it is to a proper understanding of the issue.

Those interested in the issue should note that in 2002 professor of public health Johan Mackenbach wrote in a survey:
"Overall these papers reinforce the idea that the evidence for a correlation between income inequality and the health of the population is slowly dissipating. There is very little confirmation of such a relation outside the United States. Within the United States it has still to be convincingly demonstrated that it is not due to curvilinear individual level relationships and confounding."

(Income inequality and population health, British Medical Journal, 2002)

The general public - the target audience for The Spirit Level - cannot be expected to be aware of the state of research in the field. Wilkinson and Pickett exploit the trust of their readers and give them the impression that their claims represent consensus science, when the opposite is closer to the truth.

We also note that Wilkinson and Pickett write that "there are now over 200 studies testing [the relationship between inequality and life expectancy]." They give a similar response to another one of our questions, referring to their 2006 review noted above. But in the review they themselves write:

"We identified 168 analyses in 155 papers reporting research findings on the association between income distribution and population health"
(Income inequality and population health: a review and explanation of the evidence, Social Science & Medicine, 2006)

It is unclear how this fits with Wilkinson and Pickett’s claim of over 200 studies dealing with life expectancy and equality.

Our Original Question

Q3. Correlation is not causation. This is true both for simple relationships and with multiple variables. Do you have any studies that actually establish a relationship between life expectancy and inequality, based on exogenous variation of inequality, quasi-experiments or any other well identified source of variation?

Response from Wilkinson and Pickett

Indeed, correlation is not necessarily proof of causation. However, as epidemiologists, we are trained in researching causal relationships within an observational framework. One of our critics has admitted that "sociology has been remarkably inept at providing us with the evidence and tools to create a better society". We agree, but epidemiology has been much more successful in uncovering causal influences on population health and in pointing the way to effective public health policy. Epidemiologists have been able, within an observational framework, to show that: smoking causes lung cancer; sleep position affects babies' risk of dying; social status and social networks have a profound impact of on people's risk of chronic disease, etc...

We discuss the epidemiological criteria for the establishment of causality in our book. One example might strike readers as a useful illustration. At the end of the Second World War, the USA was a very equal country, and ranked high on population health; Japan was a very unequal country and ranked very poorly on population health. Since then, these two countries have switched their relative positions: the USA is now very unequal and ranks very low on health; Japan became much more equal and its life expectancy increased faster than any other developed country till it had the highest life expectancy in the world. See also the study by Clarkwest et al referenced in point 4 below 18. There are a number of papers dealing with changes over time, with path analysis, and with causal ordering. In addition, a number of the associations found in observational studies of humans, including biological measurements, have been supported by genuine experiments on non-human primates where social status can be manipulated while material standards are kept constant.

Our Response

Wilkinson and Pickett avoid answering our question. They only cite one article, which relates to China. (Provincial income inequality and self-reported health status in China during 1991-7, J Epidemiol Community Health, 2006).

It is surprising that this paper is considered relevant by Wilkinson and Pickett when it only has 6 citations according to Google Scholar whilst Deaton's paper, with almost 100 times that many citations, is not. Also, why do Wilkinson and Pickett suddenly turn to a study focused on China when their analysis (particularly in The Spirit Level) is focused on OECD countries?

Wilkinson and Pickett do write that there "are a number of papers" which show a causal link. Why do they not refer to these papers? We repeat the question: Do you have any studies that actually establish a relationship between life expectancy and inequality, based on exogenous variation of inequality, quasi-experiments or any other well identified method of studying variation?

Wilkinson and Pickett rely on simple correlation studies. But, if we observed a correlation between headaches and taking aspirin, we would not conclude that aspirin causes headaches. Similarly, that income distribution is occasionally found to be correlated with various social ills should not come as a surprise.

It is, in fact, difficult to imagine a social problem that does not plausibly depress the income of the impacted group in some way, thereby increasing income inequality and giving rise to a spurious correlation. Repeating the flawed methodology on numerous occasions will not magically turn correlation into causation.

In the modern social sciences, researchers tend to understand the need to conduct careful statistical analysis when dealing with complex economic phenomenon such as income inequality. Wilkinson and Pickett do not rely on statistical methods that actually separate simple correlation and causality, even though the question they are engaging in is particularly prone to complex ties between variables.

Our Original Questions

4. Your most famous claim is that "inequality kills". Yet, using OECD life expectancy data, UN life expectancy data, OECD Gini and UN Gini, with different selections of countries, in several specifications, we again and again fail to replicate your result and find any statistically significant relationship between life expectancy and inequality. Is the explanation that you have relied on cherry picking – using the exact selection of measures, countries and year where such a correlation can be shown to exist?

5. Your initial defence for the lack of a statistically significant relationship between life expectancy and inequality from OECD data was that we should look at the working age population. Do you have any further defence, given that the OECD data shows no statistically significant relationship for the population between the ages of 15 and 60?

6. If inequality (rather than poverty) is strongly related to poor health, why can we not find any statistically significant relationship between inequality and health outcomes, as measured by the OECD for 16 of 19 health variables?

Response from Wilkinson and Pickett

The claim that "inequality kills" has been made for us, in publishers' publicity and in translation. However the weight of the evidence from studies of either infant or adult mortality, among both rich and poor countries, the American states, the regions of Russia, the provinces of China, the counties of Chile and many more suggests it does. Relationships between income inequality and life expectancy have been repeatedly demonstrated since 1979 13. There are over 200 tests of this link, internationally and in the US states, and the vast majority of studies confirm the adverse impact of inequality on health.

However, after periods in which income distribution has changed rapidly, the cross-sectional international association between income inequality and life expectancy have sometimes seemed to disappear 15 – only to reappear later 16. This seems to be because there are substantial lag periods between changes in income distribution and changes in population health. Every cause of death, and death rates in every age group have different lag periods. Death rates in later life are known to be powerfully influenced by experience in early life. Taking this into account it is surprising that relationships between health and inequality have been demonstrated so many times in so many different contexts. But we accept that the inequality/health relationship is one of the weaker associations demonstrated in The Spirit Level – no doubt partly for the reasons just described. Because lag periods are much shorter for infant mortality these relationships have been more consistent as have the association among US states. However, two new pieces of evidence leave little room for doubt as to the veracity of these relationships.

One is a study published in the British Medical Journal – a meta-analysis of multi-level studies of income inequality and health. This shows unequivocally that, even after controlling for individual income or education, inequality is related to significantly higher mortality rates. The second is a study showing that US states with bigger increases in inequality between 1970 and 2000 had less improvement in life expectancy than those with smaller increases.

Our Response

It is telling that Wilkinson and Pickett are suddenly backing away from the prominent claim in their book that inequality kills. We are glad they admit that "inequality kills" is misleading, but it is interesting to note that they put the responsibility on their publishers and the translators. They should have done more to make it clear that claim does not stand up.

After all, an earlier book authored by Wilkinson was marketed with the same byline "Inequality Kills" (see the Guardian review of the book "The Impact of Inequality" in 2005).

And, as we wrote in our fourth question:
"Your most famous claim is that "inequality kills". Yet using OECD life expectancy data, UN life expectancy data, OECD Gini and UN Gini, with different selections of countries, in several specifications, we again and again fail to replicate your result and find any statistically significant relationship between life expectancy and inequality. Is the explanation that you have relied on cherry picking – using the exact selection of measures, countries and year where such a correlation can be shown to exist?"

Wilkinson and Pickett still need to explain why the relationship they show in their book only appears if a specific definition of equality, selection of countries and year is chosen.

They again claim that "over 200" studies establish a link between life expectancy and inequality, which is hard to square with their review as we mentioned earlier.

There is no response to our fifth question. Why did they tell us to look at the working age population, when no correlation exists for this group either? Perhaps most importantly, Wilkinson and Pickett do not respond to our sixth question either. Let us repeat the question: "If inequality (rather than poverty) is strongly related to poor health, why can we not find any statistically significant relationship between inequality and health outcomes as measured by the OECD for 16 of 19 health variables?"

Our Original Question

Q7. If inequality is strongly related to life expectancy, why have the countries with the highest increase in inequality witnessed, on average, higher increases in life expectancy in the last two decades according to OECD data?

Response from Wilkinson and Pickett

Researching changes in inequality and changes in outcomes is difficult, and needs careful thought about lag times. The widening of income distribution which started in the late 1970s or early 1980s; followed the spread of neo-liberal economic and political thinking from the English speaking countries to other countries. While income differences rose rapidly in the 1980s in several English speaking countries, it remained stable in a number of other developed countries before spreading to them a decade or two later. The question is which period are current improvements in health related really to – current increases in inequality or the earlier stability? So you may have the relationships exactly the wrong way round. But if you have good evidence it should be presented in a peer reviewed journal. Clarkwest and colleagues 18 have shown that states with greater changes in inequality between 1970 and 2000 have had less improvement in life expectancy than those with smaller increases.

Our Response

It is of course a very valid point that correlation is not the same as causation. But Wilkinson and Pickett base all of their analysis on correlations (many of which do not stand up to scrutiny). Their logic in this case has implications for their wider work.

Wilkinson and Pickett refer to the paper "Neo-materialist theory and the temporal relationship between income inequality and longevity change" (Social Science & Medicine, 2008). The paper has three citations according to Google Scholar so again it is difficult to see why they refer to it while ignoring the work done by Deaton (only 565 citations).

More importantly, the abstract of the paper Wilkinson and Pickett refer to says:

"Contrary to findings from prior research, analyses also reveal a strong negative association between change in inequality and change in longevity once initial levels of inequality and other state characteristics are controlled."

The paper supports the Wilkinson and Pickett theory, but in its own abstract notes that its findings disagree with the results found in other research. Unsurprisingly, Wilkinson and Pickett only refer to this paper, not those that find different results.

Our Original Question

Q8. Why do you claim that more unequal nations have less creativity, (and that Portugal is as creative as the United States), when data from the World Intellectual Patent Organization shows the opposite?

Response from Wilkinson and Pickett

We used data from the World Intellectual Patent Organisation.

This shows patents per capita for Portugal at 0.6 and the USA at 1.0. In contrast, patents per capita for Japan are 7.8 and for Sweden 30.1.

Our Response

Wilkinson and Pickett describe their source as the World Intellectual Property Organization but the link is to an amateur internet site – Nationmaster. That website gives numbers for patents granted per million people.

The preferred measure for patent intensity given by the World intellectual Property Organization is patent filings per million population, and is available directly from their website. It shows that between 1995 and 2007, the number of patent fillings per million inhabitants in Portugal rose from 8.08 to 23.57. In the US, the corresponding figures are 465.54 and 800.17 for the same years.

Looking instead at patents granted, in 2008 the Portugal total is 241 and the United States of America total is 147,154. Adjusted for population, that is about 470 grants per million people in the United States against around 23 per person in Portugal.

So the latest data – from the World Intellectual Property Organization – shows there were far more patents per person in the US as in Portugal whether measured by patents granted or filed. It appears that Wilkinson and Pickett have obtained the results they did by failing to check back to the original source.

To look up the data directly from the World Intellectual Property Organizations:

1. Go to the site http://www.wipo.int/ipstats/en/statistics/patents/
2. Click on the link "Resident patent filings per million population (1995-2007)" or "Patent grants by country of origin and by office (1995-2008)".

A researcher who finds that the US has an equal patent intensity to Portugal should realise that she or he could be using a flawed data source. They should then at least check their results against the original source (the World Intellectual Property Organization in this case) and not just a site on the web.

Our Original Question

Q9. Why do you claim that more unequal nations have more mental illnesses (and that the United Kingdom has 250% the mental illness level of Germany) when data from the World Health Organisation shows the opposite?

Response from Wilkinson and Pickett

We use W.H.O. data designed to provide comparable estimates of the prevalence of mental illness. Rather than reflecting the use of medical and psychiatric services as the data on which this question is based does, the data we used is based on the scientific collection of data using standardised diagnostic interviews administered to random samples of the population. The question reflects a confusion epidemiologists are taught to avoid at the beginning of their training. The only reason W.H.O. went to the expense of collecting the data we used is that help-seeking behaviour is a very unreliable guide to the prevalence of health problems.

Our Response

After finding such a striking difference in results between the United Kingdom and Germany, it is unfortunate that Wilkinson and Pickett did not note that on another measure, even if it is dependent on help seeking behaviour, the findings are quite different.

With such a small sample of countries the reliability of the correlation derived from the survey is clearly open to question and the numbers treated, within mostly high income modern societies, is far from irrelevant and suggests the opposite conclusion.

Our Original Question

Q10. Why do you claim that "[i]n Sweden, people don't bother to check your tickets on the train or bus" when this is obviously not the case? The American audience reading the Boston Globe might believe you, but anybody who has lived in or visited Sweden will immediately see through the deception.

Response from Wilkinson and Pickett

We have, happily, visited both Sweden and the USA several times recently so our claims are based on our own experiences in those countries. But our experiences are borne out by the evidence on trust.

Our Response

All three authors of this response live in Sweden and have done so for most of their lives. We have used Swedish trains, buses and metros for transportation thousands of times. When you go on a Swedish bus, you show your ticket or buy one. Occasionally, several inspectors board the bus and check that nobody has somehow cheated (by for example buying a ticket to stop A but instead travelling further to stop B).

When you travel with a Swedish train, your ticket will be checked. In the Stockholm metro system, significant investments have been made to stop people from jumping over the barriers without paying for a ticket. And, yet again, random checks are performed on the metro.

We are unsure how the authors of The Spirit Level have come to this conclusion.

Latest Blogs:

TaxPayers' Alliance Icon

Aid spending needs to be more transparent

4:55 PM 08, Dec 2016 Harry Fairhead

TaxPayers' Alliance Icon

The sugar tax and the public finances

6:00 AM 05, Dec 2016 Harry Fairhead

TaxPayers' Alliance Icon

Working for the taxman

6:00 AM 26, Nov 2016 Harry Fairhead

TaxPayers' Alliance Icon

Further thoughts on the Autumn Statement

4:56 PM 24, Nov 2016 James Price