The Irrelevant Evidence

Despite reviewers writing that “Sanandaji is systematic and innovative in the use of historical and empirical data in explanation of the Nordic experience,” all he really does on this key point is cite two papers that he does not describe in any detail. The papers are “The long-run history of income inequality in Denmark: Top incomes from 1870 to 2010” by Atkinson and Søgaard and “The Evolution of Top Incomes in an Egalitarian Society; Sweden, 1903-2004” by Roine and Waldenström.

In both studies, the authors use tax data to determine the pretax income shares of the top 10 percent and further subdivisions of the top 10 percent — the top 5 percent, top 1 percent, and so on. This type of evidence is irrelevant for the claims Sanandaji makes for two reasons:

1. Pretax income does not count the welfare state. As the name suggests, “pretax income” refers to the distribution of income “before” counting the taxes and transfers of the welfare state. Indeed, in the Swedish paper, the authors even go so far as to strip out taxable transfer payments like “unemployment insurance, sickness payments, and parental leave payments” in order to ensure that their definition of income excludes all welfare benefits. Needless to say, an income concept that intentionally ignores the welfare state is not going to tell you much about the effect of the welfare state. For that, you need to measure disposable income, not pretax income.
2. Top 10 percent income shares are not a complete picture of inequality. There is a whole 90 percent of people below the top 10 percent, and if you want to talk about inequality, you need to break down the distribution of their incomes as well. This is especially true when you are talking about poverty, because poverty is about how much income the bottom gets relative to the middle. It has nothing to do with top incomes. Yet Sanandaji cites this evidence as proving “low levels of poverty” precede the welfare state. The Danish study does have some data on the pretax incomes of those below the top decile that it spends a section on, but that data is mostly incomplete in early years (for example, in 1903, the tax system only touched 26 percent of Danish families).

Beyond these particular problems, Sanandaji also continually conflates the claim that 1) inequality was already low before the welfare state with the claim that 2) the Nordic countries have always been fairly equal, going so far as to quote his own brother saying that the Nordics never had “full-scale feudalism” in support of the latter contention. But the studies he cites — which, again, don’t measure the welfare state — actually show that pretax top incomes used to be extremely high in these countries, before coming down significantly in the first half of the last century (i.e. during the Great Depression and the two World Wars).

Measuring incomes right after the early-century top-income collapse and before the mid-century welfare state buildup in order to suggest the Nordic countries are just generally egalitarian is kind of silly. If anything, the more plausible story is that the welfare state locked in the one-off income collapses by imposing high taxes that prevented the rich from re-accumulating the wealth that they lost. Indeed, this is the theory offered in the Swedish paper.

It is hard to know whether Sanandaji is being intentionally deceptive in citing these studies in support of some of the claims he makes or if he genuinely does not understand how they work. He does not discuss the studies at all except to quote from the abstract of one of them.

The Lessons From Sweden

Although the Danish study is limited by sparse data and a shifting income concept, the Swedish study is surprisingly robust. Old Swedish tax data distinguishes between labor income and capital income, and it is even possible to determine how much wealth those with top incomes owned over time. This allows for a more detailed examination of precisely why top incomes changed over time in Sweden. And the lesson of that examination appears to be that you trim down top incomes by taking their wealth.

The Swedish paper finds, firstly, that the pretax income share of the top 10 percent fell dramatically during the first half of the twentieth century and that this drop was almost entirely driven by the collapse of the incomes of the top 1 percent.

After the pre-1950 fall, the pretax income shares stabilize for a bit and then start to fall again in the 1970s.

The authors of the Swedish paper decompose the incomes of the top 1 percent into capital income, business income, and labor income in order to figure out what happened during this period. They find that 78 percent of the decline in the top 1 percent’s income share between 1912 and 1935 was due to drops in capital and business income. Around 57 percent of the continued decline of the top 1 percent’s income share between 1935 and 1951 was because of drops in capital and business income.

Why did capital income plummet for the 1 percent before 1950? Mechanically speaking, there are two possible causes, both of which happen to be the case.

The first cause is the decline of capital’s share of output.

Before 1920, around 35 to 45 percent of Swedish income was being paid out to capital. By 1950, it was down around 25 percent. Capital’s share dipped again in the 1970s, which is also the second time we see the pretax income shares of the top 1 percent fall considerably.

The second cause is the decline of the wealth share of the top 1 percent. At its peak in the 1910s, the top 1 percent of Swedes (in terms of income) owned over 60 percent of the country’s wealth. By 1950, they only owned around 25 percent of the country’s wealth.

When you put the two causes together, you get a pretty clear story about what happened to the pretax incomes of Sweden’s top 1 percent before 1950: they lost over half of their share of the national capital, and the capital they had left started receiving a much lower rate of return. These two changes dramatically reduced the amount of capital income the top 1 percent received, and thereby dramatically reduced their share of total national income.

Before we get into why the 1 percent lost so much of their capital and why capital’s share declined by so much, let us pause to reflect on what this paper, which Sanandaji cites approvingly, actually teaches us about how to reduce pretax income equality. If you were using Sweden’s pre-1950 experience as a guide, what you would do is expropriate the majority of the top 1 percent’s wealth and then immiserate capital further by lowering the rate of investment return. I can hear Sanandaji right now: it’s a myth that the welfare state is what gets you equality, it’s actually the wanton destruction of the capitalist class.

Returning to our story: why did the 1 percent lose so much of their wealth during this period, and why did capital’s share decline so much?

The authors aren’t much help here, writing that “the question of why the top wealth share decreased so substantially has no obvious answer.” They do go on to speculate that “the effects of the Great Depression” and the “dramatic collapse of the industrial empire controlled by the Swedish industrialist Ivar Krueger in 1932” are probably the most important causes of the wealth decline. They also speculatively reject the idea that the World Wars played a big part since Sweden was not directly involved, but I think this is an error on their part as there are significant wealth drops during the World Wars in their dataset and affluent Swedes likely owned assets outside of their own country.

While the authors provide little guidance as to why capital’s share declined so much, we do know that one way or another, workers started getting a bigger piece of the pie during this period. I don’t think it’s too out there to suppose that the Swedish labor movement had a lot to do with that. One of the errors Sanandaji makes in his treatment of the Nordics generally is that he only discusses their welfare state, ignoring their strong labor unions, which are just as important to the Nordic model and just as heavily emphasized by advocates of that model.

The final question is: why didn’t the 1 percent gradually re-accumulate the wealth that they had lost and run inequality right back to where it was?

The authors provide two answers. First, the high taxes of the post-1950s welfare state prevented the top 1 percent from having enough disposable income to amass new piles of capital.

Second, since wealth could no longer accumulate at the top of society, it instead “accumulated in the corporate and government sector and among the rest of the population.” The accumulation of wealth in the government sector was especially dramatic in the post-1950 welfare state period.

In 1950, the Swedish government owned 14 percent of the national wealth. By 1977, it was 43 percent. Once again, Sanandaji’s reduction of the Nordic model to the welfare state blinds him to its other aspects, in this case the high level of state ownership.

A Failed Debunking

Sanandaji’s attempt to debunk the Nordic model starts with him admitting that the welfare state cuts down on inequality some and then pivots to arguing that welfare states don’t matter that much for inequality and poverty. He tries to support his claim by citing studies that neither measure welfare states nor overall inequality but show the pretax income shares of the top 10 percent. Those studies demonstrate that the Swedish way to rein in top incomes is massive wealth expropriation followed by high taxes and government ownership of capital.

Needless to say, none of this debunks anything the Left says about the Nordic countries or about how to tackle inequality generally. While you wouldn’t know it from the glowing right-wing reviews, Sanandaji’s book actually lends credence to the idea that strong labor movements, aggressive taxation, and robust build-ups of the public sector are the key to cutting down the incomes of the super rich.