Piketty and the Problem of Wealth Estimation

I’ve written before about the difficulty of talking with any certainty of the how much income or wealth the top 1%, 0.1% or the top 0.00158% of the population have. However with debates around the accuracy of Thomas Piketty’s data, this question has become more topical. Chris Giles from the FT has suggested that if Piketty had used the ONS Wealth and Assets survey data to construct his estimate of the wealth share of the top 1% then he would find wealth inequality was falling. The argument in defence of Piketty is that this ONS data only exists from 2006 and is different in type to the earlier parts of his time series of wealth inequality, therefore adjusting that data to make it more like previous data is the most accurate way to measure wealth inequality.

So what are the best data sources to use when measuring the UK wealth share? The ONS Wealth and Assets survey is undoubtedly the current best UK data on wealth inequality in its consistency at measuring wealth cross the spectrum. But it’s also probably quite wrong in its estimation of wealth at the very top. As we’ve highlighted previously it suggests that the richest 1% have a similar share of wealth to 54% of the population but this is probably an underestimate of the wealth of the top 1%.

The ONS data is contradicted by a different source of information, the Sunday Times Rich List (STRL). The STRL provides data on the 1000 richest people in the UK. It suggests that these 1000 people have £518.979 billion which is about the same as 40% of the poorest households have according to the ONS Wealth and Assets survey. Those 1000 people are approximately 0.00158% of the population[1]. The top 0.00158% of the population according to the Wealth and Assets survey only have 1.71% or approximately £162.992bn. There is a fairly large difference between what the STRL suggests and what the ONS find. If the STRL analysis could be transferred to the ONS Wealth and Assets survey without changing the overall amount of wealth the top 0.00158% would have 5.45% of the total wealth. This one small change to the very top of the income spectrum would increase the top 1% wealth share from approximately 12% to 16%[2].

The reason the ONS is so far off the STRL’s figure is most likely because there estimation of the top 0.00158% is based on just three households[3]. The sample is therefore nowhere near large enough to tell you anything truly accurate about the very richest. But we shouldn’t we expect it to, the survey is already massive and expensive with responses from 21,451 households. It over samples the top 10% to try and get a more accurate picture but far more would be needed to say anything meaningful about the absolute richest, and that would cost more public money. Whilst the discussion between Piketty and his critics is important because a good theory should be under pinned by the best data possible, it’s important to remember the limitations of that data.

Tim Stacey, Policy and Campaigns Officer.

[1] Or 0.00379% of all households it makes almost no difference to this analysis whether you consider them people or households (which further underlies the point)

[2] This is not an accurate method of calculating the 1%’s wealth, it’s illustrative of the point.

[3] Or fewer depending on weighting