Why Fiscal Progressivity Discussions Are So Muddled

Yesterday I wrote about the mistaken way that I think some commentators discuss cross-country tax progressivity. Based on OECD tables and the work of Monica Prasad, the conventional wisdom is that low-inequality countries use extremely progressive transfers rather than progressive taxes to get that way. But when you look at transfer levels in these countries broken down by income decile, you often see something like this:

That sure doesn't seem like progressive transfer spending, does it? So how can the conventional wisdom be right if the graph looks like that? Why does this graph seem on first glance to so challenge the conventional wisdom? The answer lies deep in the methodological weeds. Explaining it helps to reveal why I find these discussions to be so muddled and why I think the conveying of the conventional wisdom tends to be broadly unhelpful to normal (and often even very sophisticated) audiences.

Divide By Market Income

The way Prasad calculates tax progressivity is as follows: divide households into quintiles according to their market incomes, find the market income gini based on those quintiles, find the "taxes paid" gini according to those same quintiles, then subtract the market income gini from the taxes paid gini (i.e. get the Kakwani index). This is a somewhat complicated calculation to explain here, but what it's basically getting at is dividing taxes paid by market income across income groups.

And so, yeah, when you divide income taxes paid by market income in Finland (to keep with my example), you get something that shows flat-to-regressive income taxes:

Calculating tax rates this way runs into the same problems I identified before (specifically around tax expenditures and taxes levied on transfers), but we can leave those aside here. Take the graph at face value and observe that these taxes aren't progressive.

If you are going to calculate tax progressivity by dividing taxes paid by market income like this, then the only way to calculate transfer progressivity in a comparable way is to also divide transfer incomes by market incomes. And here's what happens when you do that:

Aha, there are those "progressive" transfers we've been promised. As we know from the first graph in this post, these deciles are actually getting pretty similar amounts of transfers, but dividing by market income (which is what the Kakwani index essentially does) makes them out to be very "progressive."

So, as long as you consistently divide everything by maket income in order to define progressivity, you can say what the conventional wisdom says: they have flat-to-regressive taxes and very progressive transfers.

Levels Only

The problem here is that, though it's perfectly fine under the Kakwani approach to say this, it doesn't actually track what we usually mean by "progressive transfers." When normal people hear that a country has a progressive transfer system, they have in mind levels. They think you mean that the country is transferring way more dollars (or in this case Euros) to the bottom than the top. But that is not what is going on.

If we only look at levels to define progressivity (rather than somewhat strangely pegging everything to market income), things look like this:

Under this levels-only approach, it's clear that transfer levels are pretty flat (the red bars are fairly even across the deciles) while tax levels are very progressive (the blue bars keep getting bigger and bigger as you go from the lowest to highest decile). So, if we talk about progressivity without dividing everything by market income, we'd say that Finland achieves its low inequality, not with progressive transfers, but with progressive taxes.


In practice, people seem to use rates when talking about income tax progressivity, but use levels when talking about transfer income progressivity. This is why I am frustrated by how the conventional wisdom is conveyed and why even sophisticated sorts found that first graph at the very top to be so at odds with it. Under our normal way of discussing things, what we'd say is Finland has flat taxes and flat benefits. But we mean totally different things by "flat" for those two uses.

Ultimately, then, I suppose this ends up being a semantic point. If you stick to the rates approach across the board (i.e. divide all your levels by market income), you end up saying Finland has flat taxes and progressive transfers. If you stick to the levels approach across the board (i.e. don't divide anything by market income), you end up saying Finland has progressive taxes and flat transfers.

As for how to best convey this understanding to a general audience, I think the levels approach makes far more sense. In my experience, what people are trying to learn from this discussion is whether low-inequality countries are heavily concentrating their transfer benefits towards the bottom half or so. And when you say they use "extremely progressive" transfers, you convey the understanding that they are doing that when they aren't. It's not transfers that vary by decile. It's taxes. Thus, it's tax differences, not transfer differences, that are causing the net transfers graph to look like this:

Treatments of this topic that fail to convey this (and I think many of them do, often because even the writer doesn't understand what's going on) darken more than they illuminate. Really, the "progressivity" discussions in general do that, making the topic far more complicated and muddled than it needs to be.

Which is especially sad because the Nordic model of inequality reduction is pretty simple: use broad-based transfers to increase everyone's gross income and balance that fiscally by levying taxes that increase with income.