Did the Thirty Glorious Years Actually Exist?

Okay, I am going for a flashy title here. I should have asked whether the Thirty Glorious were as glorious as they are meant to be. This is a question that matters in debates about both inequality and the often-bemoaned growth slowdown.

In the past (say before 1950), labor force participation was quite low (relative to today) by virtue of large family sizes and most married women not working. However, when they were at-home, these married women produced something. That something was simply not included in our national accounts. When they entered the labor force, they produced less of that something. However, since it had never been measured, we never subtracted that something from the actual output generated from their increased participation.

Even before the 1950s, this mattered considerably as growth tended to be heavily underestimated (by 0.3 percentage points from 1870 to 1890, overestimated by 0.38 points from 1890 to 1910 and by 0.06 percentage points from 1910 to 1930).This was at a time when variations between the household economy and the market economy were small. Imagine the importance of overestimates since the 1950s! In a short comment reply to Emily Skarbek last year, I pointed out that adjusting for the size of the household economy meant that 1/7th of Canada’s economic growth from 1960 to 1997 (see image below and this was before one additional surge of labor participation resulting from daycare and unemployment policy reforms).


Recently, I found an old book in my library. It is Kenneth Boulding’s Structure of a Modern EconomyIn it, he makes this exact same argument. Basically, actual output today is overestimated relative to output in the past. And there are many, many, many other articles on this. In all cases, the rate of growth is heavily reduced. In a way, that means that the Thirty Glorious are less glorious (which makes the growth stagnation argument seem more defensible).

And you know what? This is consistent with attempts to correct inequality measures. Most of the attempts made to correct inequality for age, number of workers per household, the size of household and prices, they generally increase very modestly the income growth of the bottom centiles and decrease appreciably the actual level of growth of incomes at the top. While these corrections reduce the level of inequality (and the growth thereof), they also reduce the growth rate of incomes.

Is it possible that the correction to make inequality measures more comparable over time are allow us to see the point about overestimating growth since the 1950s? It means that the Thirty Glorious aren’t that glorious (at the very least, they’re overestimated). It also means that someone who could follow some of the proposed corrections to national income accounts (generally, the best source for this is the Review of Income and Wealth) for every year since 1929 (starting date of the US national accounts which could be extended by using Kuznets’s national income measures from 1913 to 1929) could propose the “actual output” of the country and see how glorious the 1945-1975 period was. That is the work of economic historians to do!

Women and secular stagnation

As an economic historian, I’ve always had a hard time with the idea of secular stagnation. After all, one decade of slow growth is merely a blip on the twelve millenniums of economic history (I am not that interested with the pre-Neolithic history, but there is some great work to be found in archaeology journals). Hence, Robert Gordon’s arguments fall short on me.

That was until I was sparked to react to a comment by Emily Skarbek at Econlib. Overall, she is skeptical of Gordon’s claims of secular stagnation. But not for the same reasons. She claims that there are many improvements in welfare that we are not capturing through national income accounts. This is basically the same point as the one made by the great Joel Mokyr (the gold standard of economic historians).

It is true that national accounts have some large conceptual problems regarding measuring output when there are massive technological changes. Yet, all these problems don’t go in the same direction. More precisely, they don’t all lead to underestimation of growth.

My favorite example of one that leads us to overestimate growth is the one I keep giving my macroeconomics students at HEC Montreal. Assume an economy with a labor-force participation rate of 50%. Basically, only males work. All women stay at home for household chores and childcare. In that case, all measured output is male-produced output. Since national accounts don’t consider household production, all the output of women in the households of this scenario is non-existent.

Now assume a technological change causing a shift of 10% of women to the workforce at the same wage rate as men. That boosts labor participation rate to 55% and output by 5%. However, that would largely overestimate growth caused by this shift. After all, when my grandmothers were raising my parents, they were producing something. It was not worthless output. Obviously, if my grandmothers went to work, there was some net added value, but not as much as 5%. However, according to national account, the net increase in GDP is … 5%.

Obviously wrong right? Now, think of the economic history of the last 100 years. Progressively, female labor-force participation increased as marriages were delayed and family sizes were reduced. Unmarried women stayed on the market longer. Then, the introduction of new household technologies allowed some married women to join the labor force more actively. Progressively, women accumulated more human capital and became more active in the labor force. So much that in many western countries, both genders have equal labor-force participation rates.

As they shifted from household production to market production, we considered that everything they did was a net added value. We never subtracted the value of what was produced before. Don’t get me wrong, I am happy that women work instead of toiling inside a household to handwash dirty clothes. Yet, it would be both statistically incorrect and morally insulting to say that what women did in the household had no value whatsoever. 

The role of household production in reducing the quality of growth estimates goes back to the 1870s! A 1996 article in Feminist Economics (which I use a lot in my own national account sections of macroeconomics classes) shows the following changes in growth rates when we account for the value of household production. Instead of increasing to 1910 and then falling to 1930, growth in the United States falls to 1930. While the growth rates remain appreciable, they nonetheless indicate a massively different interpretation of American economic history.



Sadly, I do not possess a continuation of such estimates to later points in time for the United States. I know there is an article by the brilliant Valerie Ramey in the Journal of Economic History, but I am not sure how to compute this to reflect changes in overall output. I intend to try to find them for a short piece I want to submit later in 2016. Yet, I do have estimates for my home country of Canada. Combining a 1979 paper in the Review of Income and Wealth with a working paper from Statistics Canada, it seems that the value of household production falls from 45% of GNP in 1961 to 33% in 1998. When we adjust GDP per capita to consider the changes in household work in Canada, the growth path remains positive, but it is less impressive.


I am not saying that Gordon is right to say that growth is over. I am saying that the accounting problems don’t all go in the direction of invalidating him. In fact, if my point is correct, proper corrections would reduce growth rates dramatically for the period of 1945 to 1975 and less so for the period that followed. This may indicate that “slow growth” was with us for most of the post-war era. That’s why I reacted to the blog post of Skarbek.

It also allows me to say the thing that is the best buzz-kill for economics students: national accounting matters!