Transcript Discussion of Oulton and Srinivasan (O-S)

Discussion of Oulton and
Srinivasan (O-S) on Capital
and Depreciation
Robert J. Gordon
Northwestern University and CEPR
ECB/CEPR/BdE Conference on Prices, Productivity, and
Growth
Madrid, October 17, 2003
The Macro Context of this Topic





Capital Stocks, Capital Services, and
Depreciation
Services/Stock = Utilization
Y = F(N,UK) not = F(N,K)
A leading “explanation” (really a
byproduct) of the puzzle of procyclical
Solow residual and SRIRL
Work by Basu-Fernald and Eichenbaum
Macro Error in ignoring fluctuations
in capital (capacity) utilization



Macroeconomists in the Keynesian Tradition
have always treated capacity utilization as a
byproduct of business cycles; utilization
completely endogenous
Much of the recent macro literature, esp. the
RBC and its variants, err by ignoring utilization
and then can’t explain impulse-response after
technology shocks
Steel in U. S.: 18% utilization rate in 1932
Sometimes We May Forget How
Much Utilization Can Vary
Steel Industry Capacity and Production in Tons, 1929-41
90
80
70
60
50
Actual
Capacity
40
30
20
10
0
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
What Did We Know about This
Topic in 1967?

Perpetual Inventory Capital Stocks

Fails to Reflect Cycles of Investment


Exaggerates importance of physical depreciation



Office buildings are torn down when new buildings are
constructed, same for personal computers
One-hoss shay vs. geometric depreciation
Ignores distinction between depreciation and
retirement
Physical depreciation can be offset by
maintenance, rarely reflected in official data

The story of this afternoon, rental shelter prices
Summary of Paper


Correctly endorses physical interpretation of
capital stock, services, depreciation
Why we don’t want a wealth concept



Central concept of “VICS” – volume index
Most surprising result: “no tendency for agg
depreciation rate to rise over last two decades”



asset prices – Tobin’s q, stock market bubbles
Why? Echoes Oliner-Sichel (1994) and Sichel (1997)
that “computers are just too small to matter.”
Changing share offset by declining relative price
Yikes! What is going on with U. S. productivity?
The Basic Flaw in the Paper is the
Failure to Define the Question


Are we talking about capital as wealth or as an
input into the production function (Y = F(N,UK)
Endorses geometric depreciation




Differs from my one-hoss shay approach
My productivity in last night’s hotel room
The irony of the paper is that they reject wealth
criteria but then implicitly impose a wealth
criterion when endorsing geometric depreciation
The right criterion: one-hoss shay with a
maintenance adjustment
A Misleading Message about
Aggregation


“The Main Difference between the VICS
and wealth-type measures of capital is the
way in which different types and ages of
assets are aggregated together.”
NO!

The main difference is that VICS measures
are impervious to asset-price bubbles but
wealth-type measures of capital are not.
Why Should We Care?


As Economists, because there is so much
of macroeconomics that involves TFP
Simple definitional production function




y = a + bn + (1-b)k
a = y – bn – (1-b)k
a = y-n – (1-b)(k-n)
Conclusion:

Measurement errors in k create opposite
signed errors in TFP
What the Paper Contributes

We can endorse


Physical measure of capital vs. wealth
What we cannot endorse


The theoretical section (pp. 28-39) never
writes down a production function
Paper is totally unclear vs. obsolescence-type
depreciation and its effect on wealth and the
input of capital into the production function
The section on Aggregate
Depreciation


Very important result that under chainlinking the aggregate real depreciation
rate can rise without limit, eventually
exceeding the rate on any individual asset
But “what else is new”?

We have been educated by the BEA and
others since 1996 that in a world of chainweighted deflators and real GDP, NOTHING is
additive!
US vs UK, Geometric vs.
Augmented Straight Line

Paper’s useful dichotomy


Alternative approaches to depreciation do
NOT have any impact on the growth rates of
the capital stock in steady state
But the chosen depreciation rate assumption
obviously affects the LEVEL (or VALUE) of the
stock.
The Heart of the Debate


P. 36 (“Sometimes it is argued that only
physical wear and tear should go into the
measure of depreciation used to construct
capital stocks”)
This section is totally devoid of the
question we want to answer by
constructing such capital stock measures.
Other Central Conclusions


Growth rates of wealth and of VICS are
insensitive to variations in depreciation
rates, i.e., asset lives
The LEVEL of wealth is sensitive to
depreciation rates


Why should we care? What is the question?
The input of capital into the production
function answers a very different question
than wealth in the consumption function
Estimating Depreciation in practice,
pp. 41 ff

Distinction between the “pure age effect”
and also the “effects of changing quality.”

Too much emphasis on computers




Diminishing returns to quality improvement in
computers (my anecdotes, could have done it on a
park bench)
Contrast with planes, trains, and automobiles
Here the one-hoss shay model operates with
the added impact of maintenance
Why do people buy new cars?
The bottom line, pp. 60-61


No big changes in depreciation/GDP over time
Echoes Oliner/Sichel (1994) and Sichel (1997)




Computers are still a small part of the capital stock
Structures are still a very large part of the capital
stock
Come back at 1600H for our discussion of
housing price indexes
Could you believe that housing prices is a more
important topic than computer prices?