I was the main author of a Tax Justice Network report published in 2005 called The Price of Offshore. In it we argued that $11.5trn dollars of private assets of high net worth individuals are held offshore and that the resulting tax loss was $255bn a year. Those numbers have since acquired a life of their own, being quoted, often, around the world in everything from US Senate reports onwards.
A widely cited estimate of the revenue losses due to offshore holdings of financial assets has been published by the Tax Justice Network (TJN, 2005). TJN starts with estimates of global wealth in financial assets published by Banks and Consultancy Firms (a report by Merril Lynch and Cap Gemini for 1998 and a report by Boston Consulting Group from 2003). This is combined with estimates of the share of financial assets held offshore by the Bank for International Settlement (which refers to US asset holdings, though). By combining these numbers, TJN (2005) claims that offshore holdings of financial assets are approximately US-$ 9.5 trillion. This is augmented by US-$ 2 trillion of non-financial wealth held offshore like e.g. real estate (no source is given for this number). On this basis, TJN (2005) estimates that globally approximately US-$ 11.5 trillion of assets are held offshore. Assuming an average return on these assets of 7.5 percent implies that these offshore assets yield a return of US-$ 860 billion. Moreover, the TJN assumes that these assets are taxable at 30% and thus calculates a revenue loss of US-$ 255 billion per year (in 2005).
They go on to say:
Clearly, these are rough back-of-the envelope calculations based on ad hoc assumptions on taxable rates of return and the distribution of asset holdings which, again, cannot be verified.
So much for academic objectivity.
But let’s ignore that, let’s consider the facts. First, because they failed to ask and failed to read the paper properly they got the data sources wrong. The 1998 and 2004 World Wealth Reports were used, not just the 1998 ones. And the BIS data used was not restricted to the USA.
Second, we did not combine these sources: we used them because they triangulated. That’s a perfectly acceptable statistical technique.
Third, this is not ‚Äòad-hoc’ data. It’s data from major market sources. You might call it the firm level data which elsewhere the Oxford team are so keen on, and no doubt it was prepared at considerable cost from micro sources, again of the type they seem so keen on. So what was the problem with our data?
Fourth, we based the estimate of the ‚Äòadditional’ non-financial assets or illiquid assets included in trusts on experience and information from informed persons. Actually, we’ve since been told by Colin Powell, the Chair of Jersey Finance (amongst many other things) that we seriously underestimated this. I happen to believe that such sources, when treated with care are perfectly acceptable.
Fifth, the rate of return was based on a survey of risk based portfolio returns at the time. Which for an offshore investor seemed a reasonable basis for calculating expected returns.
The tax rate was based on weighted data from the KPMG corporate tax rate survey and (rather bizarrely) data from the Forbes Tax Misery index on the highest marginal income tax rates in all the countries where we expected the owners of these funds to be really resident, with it seeming likely that the countries surveyed would be the main sources.
And we then allowed for tax withholding on part of returns and for part of the holdings to be declared. The Oxford paper ignores this and argues in the paper that this could be a weakness in the methodology. It is not: we allowed for it.
So is this a back of the envelope calculation? Clearly not. Not unless you dismiss all work by Cap Gemini, Boston Consulting, Merrill Lynch, the BIS, Forbes and others as ‚Äòback of an envelope’. Not unless you say that tax rate data cannot be used at face value. Not unless you say triangulation has no relevance. Not unless you assume it is not possible to calculate expected average returns, or that those that are published bear no relationship to real returns.
Of course you can assume all those things because the data used does not come from the usual sources an economist uses. That’s true. But then, tax havens don’t publish much data, at all. So what else were we to do?
The fact is we used verifiable sources and explicit (not ad hoc) assumptions to create a best estimate of a recognised phenomena (and surely the existence of offshore funds is not denied by the Oxford team?) to then publish a best estimate based on that work.
What is so different in that process form the work of the average economist, who uses inherently unreliable data (like published accounts) to which gross simplifying assumptions are applied to make them manipulable (for example, with regard to exchange rates, or that they are all prepared on the same accounting basis or that the tax charge on the face of the profit and loss account represents the liability due) to which a regression analysis is then applied to which arbitrary weighting of percentage significance is then attributed which is then used to provide an estimated result?
In both cases the result is an estimate. But why is ours a ‚Äòback of the envelope’ job and the alternative not? An answer is needed, because it is not apparent that a good one is available – and the Oxford paper certainly does not provide it, especially when the opportunity to question me about the work was foregone and it does not seem that the paper was read with any particular care.
All of which lends support to my opinion that the Oxford report is seriously flawed.