Tax incentives cost $138 billion…?

#taxpaysfor my PhDCongratulations to ActionAid on the launch of its new Tax Power campaign – an impressively internationalised version of the work ActionAid UK has been doing for five years now. I love the gallery of #taxpaysfor photos.

As part of the campaign launch, ActionAid asked me to help them come up with an estimate for the revenue foregone by governments in developing countries through corporate tax incentives. As the campaign briefing says, there is mounting evidence that such incentives are often ineffective at attracting the kind of investment that leads to sustainable economic growth. (This is distinct from the general rate of corporation tax, which is a whole other debate…) Certainly they are rarely put in place with any kind of cost-benefit analysis, which is why there’s so little reliable data out there.

We decided to come up with a ‘ballpark’ average figure for the revenue foregone as a share of GDP, and apply this to the total GDP of all developing countries. The scaling up part is obviously quite a simple approach, but I was quite pleased with the way we arrived at the average figure to begin with, so I thought I’d share it.

Data on ‘tax expenditures’ – that’s the revenue lost through tax incentives – is quite sparse, and where it does exist it’s plagued with inconsistencies. After quite a lot of hunting around, I managed to find about 20 developing countries where the government had published tax expenditure data, either directly or via a civil society organisation. I took the most recent year I could find in each case. I was particularly proud to have dug up a figure for Bhutan!

Tax expenditure reports can include the cost of everything from VAT exemptions to free trade zones, so it was essential to a) find something consistent and b) focus only on the kinds of expenditures that ActionAid is campaigning on. I’ve seen a few organisations cite massive figures for the cost of tax exemptions in a country where, if you go to the original source, you see that most of these go directly to ordinary people, not multinational companies. An IMF paper [pdf] says that tax expenditures probably amount to a couple of percent of GDP, but that refers to all types of exemption.

Although a few countries give one aggregate figure for direct taxes, which annoyingly makes personal and corporate income tax indistinguishable, there were 16 where I could find, or at least make a good guess at, the share of tax expenditures coming from corporate income tax (sometimes that involved running line-by-line through an itemised expenditure). So that’s the figure I used.

I ignored all other taxes from which companies get exemptions. I also excluded the expenditure on deferrals (i.e. accelerated depreciation) because in theory at least this just creates a timing difference – I presumed that in any given year the government foregoes some revenue in this way, but also receives some extra because of past deferrals. Maybe other tax brains out there can tell me if that was the right thing to do!

The data after all this processing is given in this Google spreadsheet.

I used some whizzy regression software from the LSE to check whether there was any connection between the proportion of revenue foregone and the amount raised, or GDP per capita, or the size of an economy, but I couldn’t find any meaningful relationship. That’s why it seemed like a simple average, 0.60% of GDP, would be the best way to go. Interestingly it’s pretty much the same as the figure for India, which is also by far the biggest economy in the sample.

When you think about it, rough though it is, that’s a huge ‘ballpark’ to be in.


  1. Is there a correlation in legal systems? My anecdotal hunch is that Common Law based setups (eg UK, USA) rely more on tax as a policy weapon (=> tax expenditures) than civil law codes, eg France, Germany. Effectiveness of VATs shows a correlation with legal/linguistic background (IMF Working paper WP/12/220 refers pp6,7); does tax expenditures follow the same bias?

    1. Interesting. In this case we’d be looking at colonial heritage. In the IMF paper you’re taking about, the finding seemed to be that there was an association with colonial heritage, but the actual difference was the opposite of that between the British/French systems. In any event, the sample has anglophone, francophone, hispanophone and lusophone countries, and I can’t see an obvious connection with the amounts of revenue foregone.

  2. One thing I couldn’t find anywhere in the report was any suggestion of why the tax incentives are there in the first place.

    According to the report there is a lot to be gained by removing them, but no apparent downsides. So turning that round, providing the tax incentives seems to harm the various countries for no gain.

    The closest the report comes to addressing this seems to be where it says that they are designed to encourage investment, but in practice any investment would have happened anyway. This suggests that they’re not fit for purpose, but without knowing whatt he purpose it it’s hard for the reader to judge that.

    It may of course be that this is actually the whole story, but at the moment the report to me reads in a slightly one-sided way: it gives the cons of incentives, but not the pros. A bit of historical background might go a fair way to balancing things up, without diluting the message.

  3. Martin

    This looks like a good piece of detective work and careful and reasonable calculation, given the limited evidence (and thanks for showing your workings- I wish this was standard practice).

    As is often the case though, big numbers like this, when they get a life of their own, can end up hiding the important detail, and if not carefully presented give people a misleading picture.

    Most people watching the video in which the 138 bn figure features would probably conclude that a large part of the big number relates to Africa – given the location of the one case study, the animation illlustrations, and the pronounced accent of the narrator (and all the other tax campaign reports than emphasize Africa). The impression that the issue and the big number are ‘about Africa’ is also encouraged by the cover photo of the campaign report and the examples that appear in it – Zambia, Ghana, Rwanda, Malawi. The only non-African country that gets a mention is low income Bangladesh.

    But what the small print says is that 75% of the 138 billion figure relates to upper middle income countries – China, Brazil, Mexico, Turkey, Thailand etc (unsurprisingly when you think about it, given that it is an estimate based on % of GDP). 20% relates to lower middle income countries – India, Indonesia, Egypt, Phillipines etc… and about 2% relates to low income countries. Around 5% relates to sub-saharan Africa.

    It would give a clearer impression of what the big number relates to if the video had used a Chinese narrator and the report had included Indian or Brazilian examples.

    The idea in the video that 138 billion saved through ending tax breaks across all developing countries could be used to pay for closing the gap in primary education, maternal and child health, nutrition and agriculture with 25 billion left over is attractive – but surely it implies that this would involve large transfers from China, Brazil, India, Mexico, Indonesia, Turkey, Iran, Argentina, Thailand etc… to governments in other countries for solving development challenges?

    I think the principle advocated in the campaign of governments being more transparent about the costs and benefits of tax incentives is a good one. Government policies need to reconcile domestic revenue collection and industrial policy goals. Public debate is needed to assess costs and benefits robustly at a national level. But this can only happen if the kind of bait-and-switch used in the video between costs in one place and unrelated benefits is avoided (or called out by eagle-eyed civil society). Instead it is being encouraged.

Comments are closed.