Partial preview of the text
Download Why Do Some Countries Produce So Much More Output Per Worker Than Others? | GERMAN 0270 and more Papers German Philology in PDF only on Docsity!
Hall & Jones, 1999
”Why Do Some Countries Produce So Much
More Output Per Worker Than Others?”
David Lagakos
UCLA
December 17, 2006
Outline of Paper
I Cross-Country output per worker (Y/L) accounting
I Human capital per worker
I Physical capital per worker
I TFP
I Empirical explanation of Y/L differences
I Hypothesis: ”Social infrastructure” is the fundamental cause of Y/L
level
I Empirical test of hypothesis using instrumental variables
Output Per Worker
I Output per worker yi ≡ Yi/Li
yi =
Ki
Yi
hiAi
where hi ≡ Hi/Li is human capital per worker
Data Construction
I Yi, Li, investment - Penn World Tables
I Ki - investment data & perpetual inventory method
I φ′(Ei) taken from micro studies (Psacharopoulos, 1994)
I 1-4 yrs of education: 13.4% annual return
I 5-8 yrs: 10.1%
I 9+ yrs: 6.8%
I Ai - residual
Y/L and TFP
What is ”Social Infrastructure?”
I Institutions and Government that
I Encourages productive activity
I Discourages diversion of resources
I Discourages rent-seeking activities
I Protects property rights
I Background - Kreuger (1974), Olson (1965,1982), Baumol (1990),
North (1990), et al
Y/L and Social Infrastructure
Hypothesized Structural Model
I Social Infrastructure
Si = γ + δ log Yi/Li + Xiθ + ηi
where Si is social infrastructure, Xi are included exogenous
variables, ηi is an error term.
I Output per Worker
log Yi/Li = α + βSi + ≤i
where ≤i is an error term.
I Note that we observe S˜i = Si + νi b/c of measurement error
Instrumental Variable Regression Results
Reduced-Form Regressions
Explanatory Power of Variation in Social Infrastructure
Robustness Analysis
Critique
I Instruments do not seem valid
I Hard to believe that climate played substantial role in colonization
decisions of settlers
I Extent to which Spanish etc is spoken today is randomly assigned to
countries?
I Don’t see how this constitutes exogenous variation
I Specifically, the first-stage regression seems invalid
I Can’t see any theoretical justification for a separate ’English’
instrument
I Skeptical that causation has been sorted out here
I Sympathetic with authors that deficient institutions can keep Y/L
levels low
