Two Problems With the SAT Score

Caplan, nudged by Garret Jones, <a href="http://econlog.econlib.org/archives/2016/01/econlog_reading.html">has begun a reading club reviewing the determinants of long-run growth from a historical perspective. This school of thought, which I believe Jones has referred to as the Deep History Theory of Economic Growth, takes a parameter called the SAT score as somewhat critical. This article reviews some of my problems with that operationalization.

To be clear, this has nothing to do with <a href="https://en.wikipedia.org/wiki/SAT">Scholastic Assessment Test scores.

Before I get to the critique and proposed solutions, let me discuss, and indeed validate, the general purpose of a historical perspective on growth:

1. A historical perspective on growth matters when we discuss mass migration of one society into another.
2. In discussions on mass immigration, traditional research focuses on short-run effects. There is some discussion on the long-run, but it is generally in the elementary terms of the benefits of labor growth and consumer demand, ignoring more complex effects.
3. Traditional conclusions are plausibly dwarfed or even inverted by another consideration, the long-run socio-political effect of immigration, and particularly mass migration.
4. Following from 1, 2, and 3, the historical perspective allows us to make better informed policy decisions regarding immigration.
5. Outside of the immigration discussion, economics is in desperate need of better long-run forecasting tools and methods. Economists make the strongest claims in the long-run, but they have the weakest supporting statistical data about the long-run. This historic approach may allow economists to make long-run forecasts with better statistical precision.
   1. Unfortunately, statistics follows the saying \"garbage in, garbage out.\" While better long-run forecasting is attractive in theory, the historical approach will serve only as another thorn in the side of economists if the operationalization is done incorrectly, which would lead to worse forecasts rather than better ones. That's why this article here matters.

The SAT score is an operationalization of Deep History Theory. That is, the SAT score is a kind of data which can be used to implement, test, and judge DHT. The SAT score may be defined this way:

Countries now inhabited by the descendants of historically advanced civilizations do much better than countries now inhabited by descendants of historically backwards civilizations.  How do they measure \"advanced\" and \"backward\"?  Several ways, especially state history (S), dawn of agriculture (A), and technology in 1500 AD (T).

My issues with the SAT score:

1. Technology at 1500 AD is not deep history, it is recent history.
   1. A true deep history theory should spread back millions of years to the earliest known time of mankind, or at least as early as we have good data.
   2. It's true that we basically only have bones and natural data of limited use if we go back hundreds of thousands of years,
   3. But we have useful socio-economic data describing the ancient world: Greece, China, America, and India, as well as a few other Cradles of Civilization. Good data is hard to find before 3000 BC, but it is hard not to find by 1000 BC, where we even have good data on many civilizations other than the Cradles, such as Israel.
   4. The objection to deep history is much more than a semantic problem. I am arguing that the previous implementation of the SAT score is fundamentally a bad operationalization.
   5. Putterman and Weil appeal to the Neolithic Revolution at one. That's real deep history! The Neolithic Revolution began earlier than 10,000 BC, so why are we using 1500 AD for the SAT score? There's some strong out-of-sample risk here for any conclusions drawn.
2. Technology in 1500 AD is a non-useful magic number.
   1. It would be much better for econometrics to have the entity described in this parameter as trend data instead of a point value at a relatively random date.
   2. A more sound approach, in my opinion, would be to identify the starting point and find at least 3-5 evenly dispersed dates at which to sample the technology level.
   3. 3-5 dates allows us to weakly capture the curvature of the trend data.
   4. Ideally, we would have a large sample of years with which to construct some panel data. Perhaps 50-100+ dates, and perhaps randomly rather than evenly disbursed in that case.
3. Technology is poorly measured.
   1. sss
4. Civilization is poorly measured.
   1. In the article cited above, Putterman and Weil state \"For these reasons, whenever possible we have used genetic evidence as the basis for dividing the ancestry of modern mixed groups that account for large fractions of their country's population.\"
      1. Then, perhaps worse, they state, \"By 'large' we mean 30% or greater.\" So we are perhaps ignoring up to 70% of the information contained in the biological diversity of any particular civilization's population.
      2. They say \"In cases where genetic evidence on the ancestry of mixed groups was not available, we relied on textual accounts and/or generalizations...\" That is not a very rigorous solution.
      3. Don't get me wrong, I don't fault them for using what is available, I just mean when we shouldn't be terribly confident in any conclusions drawn from this approach.
   2. The problem is that most civilizations, particularly those with advanced technology, have very usually had mixed ethnic makeup. American and Roman civilizations make for the most obvious examples, but did you know that ancient Egypt and Babylonia were ethnically diverse? You might be tempted to think that ancient China and India form exceptions, but not so.
      1. Put another way, modern biologists can't figure out how to genetically define a modern Englishman, so how could we possibly define an ancient a representative ancient American?
   3. This creates three problems:
      1. Orphaned ethnicity problem: The Tibetan People are ostensibly from Tibet, which is in China, right? However, India has always had significant numbers of Tibetans. The real problem is that neither the representative genetic makeup of China nor India looks Tibetan. What's a poor orphaned Tibetan to do?
      2. Orphaned civilization problem: Just as you can have an ethnicity with no counterpart civilization under this scheme, so you can also have a civilization with no counterpart ethnicity. What is the genetic profile of an ancient Roman anyway? Is it pretty much the genetic profile of an Italian? Nope. Rome was of course the geographic source of the ancient Roman empire, but if you look at the ancient Roman's citizens over time and found the median or mean citizen, he would not be ethnically Italian. They were the source but ultimately a minority.
      3. Ironically, the Romans were historically opposed to the northern Germans, but their genetics would never reveal this socio-political issue in history. Since the SAT score is meant to capture that very thing, socio-political and historical issues, it seems we have run into an issue in method.
      4. The third issue is that mixing matters. Mestizos are of mixed European and Native American ancestry. Perhaps they are more than 30% of each. How would Putterman and Weil treat this category? In either case, whether they treated them as essentially European or essentially Native American, they would be grossly imprecise and basically wrong. Mestizos have had a unique socio-political and historical history compared to either the Europeans and the Native Americans.
         1. Of course, Mestizos weren't really around in 1500, but this is just a convient way to dodge the issue. Mixing is a general event, not just one affecting Latin America.