10-31-2016, 07:49 AM
(10-30-2016, 04:44 PM)Mikebert Wrote:(10-16-2016, 06:47 PM)Warren Dew Wrote:(10-16-2016, 04:48 PM)Mikebert Wrote:(10-15-2016, 05:40 PM)Warren Dew Wrote: I have Ultra Society in my Amazon cart. I'm skeptical of a Turchin analysis of inequality because the inequality curves he cites in his blog don't look anything like the ones that I've seen or like the one you posted. I'll look at reviews of Ages of Discord, though - and I'd be interested in any comments you have on the points I bring up.
Turchin does not consider slavery in his inequality analysis. He uses relative wage (w) defined as median wage divided by GDP as his key measure. GDPpc is a measure of the amount of output per person. Wage is a measure of the fraction of that output that goes to workers.
My analysis uses a more sophisticated relative wage. I adjust my relative wage to account for changes in the number of people in paid employment. That is, I average in zero wages for unpaid workers (slaves) and for the unemployed. Thus, when slaves were freed, 14% of the population became wage workers. What this does is show a spike in inequality around the civil war, that does not show up in Turchin's analysis. Tuchin's analysis shows downward spike in inequality (rising relative wage) in the early 30's because GDPpc (the denominator) collapsed faster than wage (numerator). It's and artifact. My analysis shows a spike in inequality from 1929 to 1932 because bad a the Depression was for rich people, surely it was worse for the ordinary working stiff.
I also attempt to account or government transfer payments (Social Security/Medicare) in my relative wage, by adding the employer contributions to these (and private pensions) to worker compensation. Now accounted for are government transfer programs for the poor (food stamps, Medicaid, EITC, etc). However, I average in an independent measure of the top 1% income share as a second mesure of inequality, which helps account for this.
Thanks. Ignoring slavery would definitely affect things. Slaves did get a small but nonzero ration of food and minimal accomodations, but that presumably wouldn't differ from zero when determining median wage, since they would all have been receiving "wages" below the median.
How does this affect the value of the book? Should it go above Ultra Society on my reading list?
I am reading Ultra Society now. They deal with two separate strands of Turchin's research. Ultra Society is about cultural evolution, multilevel selection theory specifically, and is written for a general audience.
Ages of Discord deals with Goldstone's demographic structural theory (DST), which is the theory that explains pre-industrial secular cycles. My first cliodynamics article describes how I applied it secular cycles in England. Turchin already covered these in his Secular Cycles book. I took various mathematical formulations of DST and tried to fit them explicitly to the data (we have good data for England). They worked, one of them worked quite well in that it fit the data better than a polynomial with the same number of parameters did.
Ages of Discord is Secular Cycles for post-industrial societies, using America as the example. Turchin sees only one secular cycle between the Revolutionary 4T and the Depression 4T because that is what his five measures show. He uses relative wage (which as I noted doesn't include slavery) male height, life expectancy, age at first marriage and percent foreign born as indicators. All five show a single cycle over the 1780-1930 period.
He also presents a model for real wages than explains why real wages have stopped rising in accordance with per capita GDP after the 1970's. Basically, it is an imbalance between labor supply and demand. In the 1970's the influx of the larger Boomer generation, women and rising immigration boosted labor supply faster than demand, depressing wages. He shows this by using GDP divided by labor productivity as a measure of demand. For supply he uses labor force. It is a very plausible model. The problem is, it's garbage. Turchin believes that the BLS labor productivity measure is an independent assessment of the productivity of all workers in the economy. It's not. Actually it only covers about 75% of workers. Left out are workers for whom productivity cannot be assessed, such as persons involved in general government. These people are assumed to have zero productivity. All the GDP is assigned to those 75% for whom clear connections between work input and outputs can be made.
It's an honest mistake. But this means the percent foreign born isn't a valid measure. Then I note that the trends in male height can also be explained by trends in urbanization. Just because male height declined from those born in 1830 to those born in 1890 doesn't mean that living standards declined. The rural folks, who remained tall, were actually poorer than their city cousins, which is why so many left the farm and went to the cities. It's just that cities were (and had always been) very unhealthy places to live until the installation of sanitation infrastructure. Since such infrastructure is expensive, this did not happen until real wages rose to a higher level. By 1930, they had. So I used a simple model to explain height trends as a function of real wage and urbanization and showed it can explain the 1830-1890 decline in height too. Having an alternate explanation for height trends, means you cannot use it as a proxy for inequality anymore. The same is true for other health-related quantities like life expectancy.
This leaves just age at first marriage. In this case the 19th century data are for women only. Such marriage trends for men would certainly reflect social optimism because until recent times, men did not marry until they felt capable of supporting a family and this usually involved delay. During tough times the delay would be large, during good times less, so age of marriage would be a good indicator of social optimism. But marriage age for women are complicated by trends in women's status. This added confounding variable makes female age at first marriage unsuitable as an inequality marker. And so we are down to relative wage, which as I have said, doesn't include slavery.
But given the fact that he had *five* trends that aligned, Turchin felt fairly confident. So would I.
Now, this whole issue is a tiny part of the book, and it largely irrelevant to the main takeaway from DST. That is crises arising from rising inequality (i.e. 4Ts) are *caused* by elite proliferation. This is core DST and I believe is entirely valid. For that alone the book is worth your time. 90% of the book is sound. Turchin provides handy formulas for calculating proxies for elite proliferation. If you use these with MY inequality measure you get these secular cycle phases:
Phase First Cycle Second Cycle Third Cycle
Growth 1789-1844 1871-1890 1945-1990
Stagflation 1844-1854 1890-1907 1990-2006
Crisis 1854-1864 1907-1929
Depression 1864-1871 1929-1945
I think you can see the correspondence to turnings.
I ended up ordering Age of Discord too.
With respect to your comments, I have a few immediate thoughts.
Use of BLS statistics for productivity isn't invalid if they're the best available. It seems to me extremely unlikely that opposing trends in 25% of the population would be enough to reverse trends for 75% of the population. However, I'm surprised if BLS statistics go back far enough to differentiate between an 80 year cycle and Turchin's longer cycle. Do they?
Also, do they include slaves? Do Turchin's immigration statistics include slaves? Do his height statistics?
I am very skeptical of use of male heights only, unless that was also a limitation of available statistics. Does he explain this at all, and is there any data on overall average height or female average height?
I'd be very surprised if overall average height didn't correspond to nutritional well being. In an industrial age, however, nutritional well being is not the only form of well being any more. It would be interesting to see how well IQ corresponds to sex adjusted height, but I'm pretty sure IQ tests haven't been around long enough to differentiate the cycle lengths.