Created 12/28/1996
Go to
Brad De Long's Home Page

A Decade of the New Growth Economics

A Comment on Sala-i-Martin, "I Just Ran Two Million Regressions"

J. Bradford De Long

University of California at Berkeley and NBER

December 1996

Let me focus on Xavier Sala-i-Martin, and note that examining Hall and Jones leads basically to the same conclusions.

Here we have Sala-i-Martin's attempt to write a compressed twelve-page Summa Theologica of the past decade's work in empirical studies of cross-country growth patterns. Whatever cross-country regression of rates of growth on political and economic variables any future researcher chooses to run, the chances are very good that it will be among the two million total (455,126 separate) regressions run by Sala-i-Martin.

My first thought was that Sala-i-Martin had constructed the economist's equivalent of Jorge Luis Borges's "Library of Babel"--the library that contains all possible books, but that its useless because nothing can be found. Yet that is not correct, for Xavier Sala-i-Martin has managed to provide us all with a rough synopsis of the results of all his regressions.

An economy has probably grown rapidly if, in the post-WWII era, it began with a relatively low level of GDP per worker (hence lots of room for catch-up growth); had a relatively high life expectancy and primary school enrollment rate (hence lots of human capital); had high machinery investment (hence much learning-by-doing and acquisition of embodied machine-age technologies) and non-machinery investment (hence a high rate of capital accumulation; was committed to free trade; had a Confucian, Buddhist, or Islamic tradition; had strongly established the rule of law, political rights, and civil liberties; had strongly established capitalism; had a relatively high share of GDP in the mining sector; and was located relatively far from the equator.

An economy has probably stagnated in the post-WWII era if it is an ex-Spanish colony that had a Protestant or Catholic religious tradition; had high exchange rate distortions and a volatile black-market exchange rate premium; had a high primary-sector share of its total exports; was involved in wars, revolutions, and coups; and was located in Latin America or sub-Saharan Africa.

You may yawn, and say that you knew all this. But let me tell you some other things that you knew--that you thought you knew-- that turn out to be false:

So I finished the paper thinking that we have actually learned a lot from studying post-WWII patterns of cross-country economic growth: that the number of supposed truisms that actually turn out to be true is no greater than the number of supposed truisms that turn out to have been false. We actually know something, now, about the prerequisites for successful industrialization--and are in a position to give much better advice to developing economies than economists were fifty years ago.

Now do economists have anything to say to developed economies? Developed industrial economies already have civil, political, economic, and human rights; efficient governments; low rates of inflation; open economies; plentiful supplies of entrepreneurs; and high rates of investment in human capital. Some economists--like Robert Barro--think that the advanced industrial economies already have things about as good as they get, and can look forward to nothing more than 1.5-2.0 percent per year output per worker growth in the future.

I am not so sure, principally because I have always been powerfully impressed by the correlations between investment and growth. It's possible that investment is high where growth is expected to be high. But it's also possible--and seems to me more likely--that the theoretical tradition associated with Paul Romer has identified potentially important sources of externalities: strong potential links between various kind of accumulation and economy-wide total factor productivity growth. In 1991 Larry Summers and I argued that no matter what produced a high rate of investment in machinery and equipment--whether it came from a low price of capital goods, from a high savings rate, or from an investment pattern tilted toward machinery--the apparent social returns to investments in machinery and equipment are very high indeed.

Now Xavier Sala-i-Martin comes, saying that he has searched over 58 other variables in groups of three and found virtually no alternative combinations that can explain away a high growth-machinery connection.

So I think that we economists may well have something to say to developed economies interested in spurring economic growth as well.

		                              (1)	    (2)       (3)
			                                  Standard	CDF
		                              Beta	Deviation Non-Normal
*     Level of Income in 1960
*     Life Expectancy in 1960
*     Primary School Enrollment in 1960
1	High Equipment Investment	   0.2175	0.0408	1.000
2	Free Trade                  	0.0195	0.0042	1.000
3	Confucianism          	      0.0676	0.0149	1.000
4	Rule of Law	                 0.0190	0.0049	1.000
5	Strength of Islam           	0.0142	0.0035	1.000
6	Established Political Rights     0.0026	0.0009	0.998
7	In Latin America           	-0.0115	0.0029	0.998
8	In Sub-Saharan Africa      	-0.0121	0.0032	0.997
9	Established Civil Liberties	 0.0029	0.0010	0.997
10    Revolutions and Coups	      -0.0118	0.0045	0.995
11    Fraction of GDP in Mining 	  0.0353	0.0138	0.994
12    Black Market Exchange Premium   -0.0290	0.0118	0.993
13    Primary Sector Exports in 1970  -0.0140	0.0053	0.990
14    Strength of Capitalism      	0.0018	0.0008	0.987
15    Involved in War            	-0.0056	0.0023	0.984
16    High Non-Equipment Investment    0.0562	0.0242	0.982
17    Latitude	                    0.0002	0.0001	0.980
18    Exchange Rate Distortion        -0.0590	0.0302	0.968
19    Fraction Protestant        	-0.0129	0.0053	0.966
20    Fraction Buddhist           	0.0148	0.0076	0.964
21    Fraction Catholic          	-0.0089	0.0034	0.963
22    Former Spanish Colony	      -0.0065	0.0032	0.938


Created 12/28/1996
Go to
Brad De Long's Home Page

Associate Professor of Economics Brad De Long, 601 Evans
University of California at Berkeley; Berkeley, CA 94720-3880
(510) 643-4027 phone (510) 642-6615 fax