It is said that nothing in this world can be said to be certain, except death and taxes. Yet, we neither have valid vital statistics nor reliable information on taxes for many developing countries. While the importance of taxation for development is increasingly recognised, basic data on whether a country is collecting more taxes now than earlier, or whether country A or country B is collecting more taxes than another is actually far more difficult than one could expect.
In Poor Numbers: How We Are Misled by African Development Statistics and What to Do about It, I argued that for many countries, the quality of the statistics made it difficult to say something definite about the pace, let alone the direction of change, for vital indicators of social and economic development. I argued that this was a fundamental knowledge problem in development studies, which had been neglected. The question of validity and reliability of the basic numbers makes comparison across time and space very difficult and puts estimates of per capita trends in serious doubt.
This problem has been and continues to be particularly acute for data on taxation for development. The new dataset launched by the International Centre for Tax and Development seeks to remedy the problem.
The problem is more manageable if you are comparing one country with itself. That seems simple enough, for instance, comparing the tax receipts in 2010 with the tax receipts in 2000. However, there may have been changes in legal definitions of tax – and substantive changes in how taxes and duties are levied, and that they are therefore recorded differently – so that categories of indirect and direct taxation may not be directly comparable over time, or at the very least, requires diligent work to make sure that the time series is useful for comparison over time. The problem gets harder the further back in time you go.
These issues increase in magnitude when you want to compare levels and rates of taxation across countries. Because a nominal amount of total tax or direct tax in itself is not immediately a meaningful metric, it is usually expressed in constant terms or as a share of GDP. Here’s where the problem sneaks in.
We know that calculating GDP levels is not an exact science. The most famous recent examples include the 89 percent increases in GDP in Nigeria. Smaller adjustments that are in the order of 20 to 30 percent would still cause noise in any series that invalidates a comparison of tax as a share of GDP in a country over time or across countries. While it is possible to smoothen out the effects and/or correct for changes in definitions across time, comparisons across space of more than a few counties will invariably involve a rather big leap of faith for most, if not all data users.
Yet the most perceived pressing problem thus far has been availability. Data users would like to see global sample across the century. The new data set does a better job with availability. Borrowing data from a range of data sets, such as the OECD and using data reported in IMF surveillance reports, the compilers have succeeded in filling more gaps than have been filled before. The same way the data compilers hope they have weeded out the worst mistakes – simply by comparing GDP and Tax data from the IMF, World Bank and OECD sources, then one can hope that they ‘pick’ the better alternative (in the spirit that best should not be the enemy of better). There are still gaps, and for most countries we still do not go further back till the 1980s. Here is what I think are the major issues for future work.
- Extending the time series. If you work with a subset of countries, and use sources that are not yet entered into international datasets, but rather consult primary sources such as statistical abstracts and colonial archives, for a typical country, you may now be able to download a time series that goes back to 1980 or 1990. One key problem for teasing out interesting policy implication is that in the period 1990 to 2010, there is not so much interesting changes in the political economy, for instance, if one could draw upon data from the 1880s to the 2010s. There is a revival in African economic history, and much of the work that is being done is exactly this kind of painstaking work of unearthing and harmonizing unused historical sources.
3) 3) Write the history of taxation. Understandably the focus is on quantification, but while one is measuring how much – that answer is unintelligible without a careful study of how the fiscal institutions have evolved in less developed countries. One key historical debate – with obvious contemporary policy relevance – is whether taxation systems followed metropolitan blueprint or whether they are endogenous (which in this context means: responding to local conditions). When doing such analysis the key is to ask the right questions. Too often, work on economic growth and institution uses the ‘subtraction approach’ – where the causes of lack of development are found to be that less developed countries do not have the characteristics of the developed countries. Thus one finds that the problem with the Tanzanian taxation system is that it is not like the Norwegian tax systems. To say anything useful about tax and development therefore, one has to understand the local historical, political and geographical constraints, and figure out how to work with these.
It is curious that to date we have very weak datasets on taxation – considering its importance. I think the main reason is a conflict between validity and reliability. Validity in terms of whether the measure – like bathroom scales, is correct or not. The main compilers – which in the case of the main data source in this specific context is country authorities and the IMF surveillance team – is solely interested in getting as valid data as possible for this year, last year and a projection for next year to give immediate policy advice. The data is collected for immediate operational needs and thus validity matters. Analytical needs of scholars are different: they would like the data to be reliable: it does not matter so much if the data are incorrect, as long as the error is similar across time and space. That’s why a lot of the harmonisation of these data often is neglected – the rewards and needs are not as immediate. But that does not mean that this work is not important – on the contrary – as argued here, we need evidence to inform policy advice. The efforts of the International Centre for Tax and Development are laudable, but I will argue that in providing a historical dataset on taxation for development countries – with the necessary metadata (including the historical context which these observation are taken from) the brunt of the work will ahead of us, not behind us.
Morten Jerven is the author of several works including, Poor Numbers: How we are misled by African development statistics and what to do about it, Cornell University Press, 2013. Economic Growth and Measurement Reconsidered in Botswana, Kenya, Tanzania, and Zambia, 1965-1995, Oxford University Press, 2014, and Africa: How Economists Got it Wrong, Zed Books, 2015 and Measuring African Development: Past and Present, Routledge, 2015. Morten Jerven is an economic historian, with a PhD from the London School of Economics. Since 2009 he has been working at the School for International Studies at Simon Fraser University in Vancouver, Canada.