A Tale of Two Estimates

A brief review of Mizoguchi, et al.(2019)’s estimate of the national accounts of colonial Korea prepared in response to the request from the Maddison Project, focusing on the comparison with Cha and Kim (2012).


Mizoguchi, et al. (2019) Kankoku, kitachosen, Azia choki keizai tokei [Historical Statistics of Asia: South and North Korea], vol 4 (溝口敏行、外 アジア長期経済統計 4 韓国・北朝鮮) (Tokyo:Toyokeizaishimposha)

Cha, Myung Soo and Nak Nyeon Kim (2012) “Korea’s First Industrial Revolution,” Explorations in Economic History


There exist three key differences between Cha and Kim(2012) and Mizoguchi, et al. (2019), which refer to 1) the structure of the national accounts estimated, 2) the growth trends, and 3) the structural change under Japanese rule.

While Cha and Kim (2012) provides estimates from expenditure and production sides, Mizoguchi, et al.(2019) focuses on GDP, leaving out GDE, which has been available in previous estimates by Mizoguchi(1988). 

As Cha and Kim(2012) shows, per capita output growth took place consistently under Japanese rule without either slowing down or speeding up.  According to Mizoguchi, et al.(2019), per capita output was being reduced in the 1910s, which was followed by accelerating growth in the following two decades.  As a result, per capita output growth from 1911-40 was 2.3% (Cha and Kim(2012)) vs. 1.5% (Mizoguchi, et al.(2019)) per year.

The gap in per capita output growth arises from output, rather than population, growth estimates being divergent.  Although the two studies estimated population growth using different sources (mortality information from genealogies in Cha and Kim(2012) vs. from the colonial police in Mizoguchi, et al.(2019)), the population growth from 1911-40 are marginally different, 1.3%  (Cha and Kim(2012)) vs. 1.2% (Mizoguchi, et al.(2019)). 

Turning to structural change, both estimates indicate declining share of agricultural output and rising share of manufacturing output.  However, while the tertiary sector accounts for an increasing share of GDP in Cha and Kim(2012), Mizoguchi, et al.(2019) suggests that it contracted in terms of output share (by more than 17% points) and absolute magnitude (by nearly 40%)  in the 1910s, which was followed by expansion in the 1920s and 1930s.

The shrinking service output in the 1910s appears as counterfactual, given the evidence of population growing over 1% per year and rapid expansion of market activities driven by the improvement in communication and transportation in the 1910s. In addition, given that colonial Korea remained a small open agricultural economy, I found it implausible that incomes grew faster in the 1920s than in the preceding decade, when primary producers like colonial Korea suffered a hugely negative terms of trade shock (known as the interwar agricultural depression), than in the 1910s, when the agricultural term of trade improved sharply as European countries embarked on monetary expansion to finance WWI.

The stature growth under Japanese rule as predicted by per capita output growing 2.3% per year is reasonably close to observed height growth, which is 2.1 centimeter (Kim and Park, 2011, EEH). According to Steckel(1995, JEL), one percent growth in per capita output tends to be associated with stature growing by 0.0397 centimeter, which implies Koreans becoming taller by 2.4(=0.0397 x 2.3 x (1940-1911)) centimeters from 1911-40.

Mizoguchi, et al.(2019) provides per capita output series in terms of 1990 GK dollars, which can conveniently plugged into the Maddison database without further modification. I found it puzzling however that the income gap between Japan and Korea as suggested by the GK dollar figures was wider in 1940 than in the early 1950s, when the South Korean living standards had suffered a series of adverse supply shocks including de-colonization, the split into two Koreas, and finally the devastating Korean War from 1950-53. This issue, which has been raised by a paper by Nak Nyeon Kim and Ki-Joo Park published in Hitotsubashi Journal of Economics in 2017, exists, probably because the per capita output estimate by Mizoguchi, et al.(2019) grows too slowly.

Where did the divergence originate? The answer seems to lie in the different ways how service output, accounting for the lion’s share of the tertiary sector, has been estimated. As described in detail in N. Kim, ed. (2012) Han’guk ui kyongje songchang [The Korean Economic Growth], 1910-2010 (Seoul National University Press), service sector output in Cha and Kim(2012) represents the sum of output from different branches of service sector, which has been estimated using different types of information and following separate procedures depending on sub-sector. Mizoguchi, et al.(2019) set service output equal to the number of service labor input times the wage rate in service sector, and the contraction of service output in the 1910s is driven by the decrease in their labor input estimate by more than 30% from 1911-20.

The labor input estimate used in Mizoguchi, et al.(2019) is based on the information on “gainful employment” by sector, which included the time series of the number of part time service workers normally employed in agriculture.  Mizoguchi, et al.(2019) incorporated the part time workers in the labor input in service sector using a fixed discount factor, 0.5.  As resources were being reallocated away from agriculture in pre-1945 decades, the number of the part time service workers fell, driving the downward trend in the service sector labor input and output in the 1910s in Mizoguchi, et al.(2019).  The decline in the number of the part time service workers took place in tandem with the increase in full time service workers, suggesting occupational specialization in progress, whereby the part time workers probably earned a rising share of their incomes as service workers over the 1910s.  Hence, it seems more reasonable to me to incorporate the part time workers into service sector labor input applying an increasing, rather than a constant, discount factor (say from 0.25 to 0.75?).  Service output re-estimated using this procedure might increase in the 1910s as in the following decades.

Last, but not least, estimating the aggregate output from the two distinct perspectives carries the advantage of allowing one to check the integrity of estimates as well as to obtain additional insight on the causes driving economic growth.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s