Some interesting graphs from the WTO world trade report

I mainly work in applied trade and investment policy at the moment, and I like graphs, so here are some graphs from the latest WTO World Trade Report that tell (or could tell) interesting stories about the state of the global trade system at the moment:

Investment

Not only is the world starting to invest more of its money in developing countries, developing countries are beginning to invest more abroad as well:

merchandise

Global value chains

The big thing in trade economics at the moment is trying to get a grips on the implications of commodity chains, or global value chains, or global supply chains, or dispersed production networks (we’ve yet to settle on a definitive term, although ‘global value chains’ seems to be winning, at least among the papers I read.) This graph shows the distribution of GDP per capita for countries that participate strongly in value chains and for countries that participate only weakly. It’s not clear which direction the causality runs here (are you more likely to be part of GVCs if you’re rich, or are you more likely to be rich if you’re part of GVCs?) or even whether there’s causality at all, but it’s a good way of getting a handle on the economic geography of trade networks:

gdp value chain

Part of the reason global value chains are interesting is that it heightens the case for tariff reductions. We’ve always known that tariffs harm producers that use imported goods in making their final products, but this effect becomes even more important when participation in global value chains increases. With respect to tariff rates on parts and components, there’s been convergence between developed nation tariffs and developing nation tariffs over the past few decades:

tariffs parts and components

Around the Asia-Pacific (perhaps elsewhere as well) you hear policymakers talking about ‘climbing the value chain’, or some similar verb. Everyone wants to be up at the high value-added end, where you get paid for, say, designing a car rather than putting it together. Of course, not everyone can actually be at the high end of value chains. This shows an index of position in value chains with the 1995 position as a green diamond and the 2008 position as a blue bar.

where in the value chain

There seem to be some patterns in this data (for example, countries that were higher up the value chain in 1995 seem, on average, to be even higher in 2008) but it’s hard to tell because it’s not a particularly easy chart to read. When you have a comparison between two positions, I like to use a slope graph, which can make some of this a bit easier to see.


How many households pay net tax in Australia?

I managed to miss an article by Professor Richard Holden the other day about tax reform. Things didn’t start off well when Professor Holden advised of the need for a ‘national discussion’ about tax reform. (I’ll bring the stale biscuits and teabags; you bring the chin-strokey phrases.) Holden contends that Australia’s tax system is plenty progressive already:

First, let’s look at the baseline. Australia has a very progressive income-taxation system. The top 10 per cent of earners pay 46 per cent of total income taxes. The top 2 per cent of earners pay 26; the top 1 per cent pay nearly 18 per cent. The bottom 20 per cent pay 2.5 per cent. This makes Australia’s income-tax system among the most progressive in the world.

First of all, it’s important to know what we’re talking about. There is some confusion sometimes about what we mean by ‘progressivity’. In general, it is simply used to mean a system in which the higher your income, the more tax you pay as a proportion of your income.

But Professor Holden hasn’t talked about tax rates—he’s talked about tax paid by different income groups as a proportion of all tax paid. The two things are related, but they’re not the same thing, and you can’t use evidence about total tax paid as a proportion of the national income tax bill as evidence about the progressivity of income tax. In fact, the proportion of tax paid by an income group is a function not only of the progressivity of the tax rate schedule, but also of the distribution of pretax income.

Consider this simple, rather extreme, stylised example: Imagine there are two people in the economy: Jay Gatsby, who is rich, and Oliver Twist, who is poor. Jay Gatsby earns $1 million in a year, and Oliver Twist earns $10.  Now assume that the tax structure is as follows:

First $5 of income: 50%

Every subsequent dollar: 5%

That this is a regressive tax is pretty easy to see. (Oliver Twist’s marginal tax rate is 5%, but his average tax rate is (50% of $5 = $2.50) + (5% of $5 = 25c) divided by $10 =27.5%. His average tax rate exceeds his marginal tax rate, which is one quick way of checking if a tax is regressive.) But who pays the tax in this economy? Oliver Twist pays $2.75, while Jay Gatsby pays (50% of $5) + (5% of $999,995), which is $50002.25. Clearly Gatsby is paying the vast majority of tax —99.9945% of all tax in this economy—but it’s not because income tax is progressive; it’s because the distribution of income is so unequal.

The two concepts are related (ceteris paribus, an increase in the progressivity of a tax will see the share of tax paid by top income earners increase) but they’re not synonymous. There are indeed good reasons for supposing that the Australian system of direct taxation—as in other Anglophonic countries—is quite progressive, but not because of the share of the total income tax bill that is paid by high income earners. But my big gripe comes with another paragraph:

And income taxes are just one component of the broader taxation-benefit system. There are other taxes and benefits designed to redistribute income, such as the Family Tax Benefit, the Schoolkids Bonus and the aged pension. Then there’s a range of ostensibly free services such as healthcare. Finally, there are other taxes, such as the GST. The ABS estimates that when all of this is factored in, only the top fifth of households ranked by income pay any net tax.

I have numerous problems with this claim, which is one I’ve seen in many other places recently. First of all, let’s see where he’s getting this data from. I suspect, though he hasn’t made it clear, that what he’s looking at is the distribution of taxes, transfers and in-kind transfers from the 2009-10 household income survey. If you add up average income and ‘production’ taxes for each quintile and subtract government transfer payments and ‘in-kind payments’ like education and healthcare spending, we get something like Professor Holden’s claim: Net tax paid   Now, this is people ranked by their final income. That is to say, it ranks households by their income inclusive of welfare and education and health transfers. Then the households are split into five groups, and the average ‘net tax’ paid is computed. This tells us something interesting, but it’s not the only we could rank households. For example, we could rank households based on their privately earned income and then perform the same exercise. Doing this, we would see the following averages of ‘net tax’ paid: Screen Shot 2014-10-05 at 2.31.07 pm   Now, suddenly, we have the top two quintiles paying net tax on average, instead of just one. So the ranking of households can affect these kind of calculations somewhat dramatically. I also think that the way Professor Holden and I have been calculating this metric called ‘net tax’ is an odd way of looking at what most people would think of when they hear the phrase ‘net tax’. To me, at least, while there is something intuitive about subtracting welfare payments from tax payments to get ‘net tax’ (that is, considering a welfare payment as ‘negative tax’), it seems odd to also subtract the imputed value of health and education spending by considering it ‘negative tax’ as well. It’s also a calculation that will of its nature make it seem as though the poor are paying less ‘net tax’ than the rich, since health and education services are used quite heavily by low-income households (again, ranking by private income rather than final income): Screen Shot 2014-10-05 at 2.36.19 pm   What about other government services, like national defence, police and the judicial system? These are also government services, but they’re ones that we may have reason to believe disproportionately benefit well-off households, who, possessing more wealth, have more to lose if there were to be a breakdown in law and order. Some people might say that education is different. After all, if the government charged full cost-recovery prices at government-run schools, but gave you a tax credit for the entire amount, then you’d be paying less tax, so in a way, using a free government school is kind of like a negative tax.

To which I respond: isn’t it the same with other government services? National defence is a bit hard to do like this, but fire services, for example, used to be  run by insurance companies. What if the government decided to let the private sector deal with the problem of burglary and gave tax credits for insurance premia? You would insure your house, and people with nicer houses and better possessions would have to pay higher premia. So the fact that everyone enjoys the same access to the police now is exactly like education in its negative-tax implications, except that the benefits go to the well-off more than to the poor. So I don’t really see why we should deduct these ‘in-kind social transfers’ from tax paid to arrive at ‘net tax.’ What happens if we define net tax as tax paid minus welfare payments? Now we find that it’s only the bottom two quintiles that don’t pay ‘net tax’ on average: Screen Shot 2014-10-05 at 2.47.57 pm So definitions matter. But so too does another little phrase I’ve been sneaking into the above quietly: ‘on average’. While it’s true that, using the definition directly above, the bottom quintile of households pays –$329 a week on average of net tax as defined above, it’s also quite obvious that not every household in this group pays this amount of net tax. It’s not even clear that every household in this group pays negative amounts of net tax, or that every household in the top quintile pays positive amounts of net tax. In fact, we have every reason to believe that there are net taxpayers in the bottom quintiles as well as people who pay no net tax in the top quintiles.

But even if we assume that net tax paid is an increasing function of private income, we still encounter problems when we get to the quintiles where net tax paid is close to 0. For example, in what I assume is Professor Holden’s calculation, the fourth quintile pays on average negative $111 in net tax per week. But even if net tax paid increases as income does, the negative average in the fourth quintile might obscure quite a lot of net positive taxpayers in the same quintile. To illustrate this, I’ve simulated 100 ‘people’ each of whom pays net tax based on a uniformly random distribution between -100 and 100. I’ve then ordered the households by net tax paid and calculated the average tax for each quintile. Here’s one realisation of this simulation: Screen Shot 2014-10-05 at 1.38.47 pm The mean of the third quintile is negative, which would probably lead someone like Professor Holden to conclude that only ‘two-fifths of households pay any net tax’. But that would be wrong. As you can see, the third quintile average is composed of some household that pay negative tax and some that pay positive tax. In fact, in the above random distribution, 50% of all taxpayers pay net tax, even though on average only two quintiles (representing 40% of taxpayers) pay net tax. I also ran a Monte Carlo simulation of the above exercise. I ran the program 10,000 times and for each repetition calculated the percentage of taxpayers who paid net tax as well as the number of quintiles who paid net tax on average. I plotted a histogram and kernel density curve for the number of net taxpayers in the simulation for which there were only two quintiles that paid ‘net tax’ on average: Screen Shot 2014-10-05 at 3.27.37 pm  What this shows us, apart from some rather trivial properties of sample means, is that there are some cases where sixty per cent of households pay net tax, but only two quintiles, representing 40% of households, pay net tax on average. There are also cases where fewer than 40% of households pay net tax, but two quintiles still pay net tax on average. However, in 97.5% of cases, even though only two quintiles paid average positive net tax, more than two fifths of all taxpayers paid positive net tax. (I should emphasise that this is just using made-up data to illustrate a point.)

So when Professor Holden says “The ABS estimates that when all of this is factored in, only the top fifth of households ranked by income pay any net tax”, I say, the data can’t be used to make statements like that. What you can talk about are averages for quintiles. But even if you accept his ranking methodology and even if you assume that net tax increases with income, you can’t say how many households pay net tax just by looking at quintile averages.