SV: US TFP Growth

["Per Gunnar Berglund" <pgb@xxxxxxxxxxxxxxxxxx>]

Dear Derek Sicklen,

Your questions are directed to PKT-ers in general, but when replying I must
stress that these are my own subjective opinions. Even if I do consider
myself a Post Keynesian, I do not claim to be a representative PK-er.

My best wishes,

Per Gunnar Berglund

----------
> Från: Derek Sicklen <derek@xxxxxxxxxxxxxx>
> Till: POST-KEYNESIAN THOUGHT   <pkt@xxxxxxxxxxxxxxxx>
> Ämne: US TFP Growth
> Datum:  den 3 mars 1997 21:53
>
>
>
> A recent OECD publication (Technology and Industrial Performance, 1996)
> argues that US Total Factor Productivity - TFP - growth has increased
quite
> substantially since around 1977.
>
> TFP is defined as the "weighted average of labour and capital
productivity,
> with the period averages for factor shares used as weights" (p59).
>
> This contrasts quite sharply with trends in other measures of US
> productivity including real non-farm output per hour, which have shown
> almost no growth over the same period.
>
> Further, the use of the TFP measure shows that US productivity growth has
> been far superior to that of other G7 countries over the period.
>
> Thus the "US model" could be said to deliver better productivity growth
than
> that of those other countries.
>
> 1.  Are PKT members happy with the use of TFP?

No, I am not happy with the use of TFP.

>
> 2.  My understanding of TFP is that it requires a neoclassical production
> function approach which implies the notion of the marginal productivity
of
> capital.
>

You are perfectly right. Do read the 'Humbug production function' article
in The New Palgrave on this!

> 3.  If this is the case, didn't the Cambridge capital debates of the late
> 1960s and early 1970s effectively destroy the concept of the marginal
> productivity of capital?
>

I think so.

> This is an important issue, as the choice of productivity measure will in
> part determine one's assessment of US economic performance both in
absolute
> and relative terms. The policy implications of choosing either definition
> could well be total opposites.
>

Extremely important, I am afraid. The trouble lies in defining the 'volume
of capital' concept, as distinct from the value of capital assets --
splitting asset values in p and q dimensions, that is.

Unless you can find a method of defining that volume concept, you will
unfortunately also be unable to define properly the volume of output
concept. Therefore, not even an ordinary *labour* productivity concept can
be defined without first defining the volume of capital -- the numerator is
simply not there! Proof:

Output = Consumption + Capital formation (+ Net Exports in the open economy
case, which I skip here to simplify)

hence,

Volume of Output = Volume of Consumption + Volume of Capital formation

But if

Volume of Capital formation = (annual) Change in the Volume of Capital
Stock

it follows that

Volume of Output = Volume of Consumption + Change in the Volume of Capital
Stock

Conclusion: The volume of output cannot be measured without *first*
defining the volume of capital stock. Thus, the annual output volume will
be affected by the assumed rate of 'meshing' of old capital with new
output.

The UN System of National Accounts (SNA) recommendations take a pragmatic
stance using the Perpetual Inventory Model to write down real capital
assets. In that way, the constant-price estimates of 'gross capital
formation' are diminished by the PIM-computed 'depreciation' to estimate
the 'net capital formation' which is taken to be the (annual) net addition
to the capital stock. In this way, a capital stock volume measure can be
cumulated over a long sequence of years. The period must be long enough to
allow for the slowest-depreciating assets (buildings) to wear down to a
negligible level. In practice, this is usually some 50--70 years.

The SNA may be criticised on several grounds:

First, it does not really deal with the fundamental question of what the
'capitalness' of an object is. According to the PIM approach, a certain
part (percentage) of the stock is supposed to 'go into' or 'mesh' with the
annual output, in the same manner as the value of a stock is written down
in business accounting. This may or may not be a useful concept, but as far
as I know no serious attempts have been made to justify this concept
theoretically.

Second, since the writing-down period is very long for some objects, it is
practically impossible to obtain good estimates of depreciation
coefficients. And apart from these practical problems, it is impossible to
judge which historical information basis is objectively relevant. This
theoretical problem is similar to the adaptive expectations problem: Assume
that we are to write down a stock of automobiles. What if the 'ordinary'
lifelength of an automobile has changed over time? Suppose it was twelve
years in 1975 and fifteen years in 1995. How do we extrapolate this into
the expected life of a newly produced vehicle? And how exactly do we
estimate the shape of the so-called survival curve?

Well, Derek, as you can see this whole thing is a muddle, where pragmatic
solutions have been adopted in place of theoretically sound ones. The
latter kind is clearly preferrable, provided that they can be made
operational in a reasonably simple way. The trouble is that nobody, as yet,
did supply any of those warranted solutions. Maybe you have some
suggestions?

Let us keep the discussion of this intriguing subject going on the list!

/PGB

> Could I get some feedback please?
>
>
> Derek
>
>
>
> ______________________________________________________
>
> Derek Sicklen
> Director
> Australian Economic Analysis Pty Limited
> Kirribilli NSW 2061 Australia
> Email: derek@xxxxxxxxxxxxxx
> ______________________________________________________
>