You are on page 1of 3

Economics Letters 114 (2012) 91–93

Contents lists available at SciVerse ScienceDirect

Economics Letters
journal homepage: www.elsevier.com/locate/ecolet

Variable factor shares, measurement and growth accounting


Hernando Zuleta ∗
American University in Bulgaria and Universidad del Rosario, Balkanski Academic Center, 54 Alexander Stambolinsky St. Blagoevgrad 2700, Bulgaria

article info abstract


Article history: Recent evidence shows that factor shares are not constant. Consequently, growth accounting exercises
Received 14 September 2010 rely on a false assumption and a measurement problem arises. We propose an empirical methodology to
Received in revised form solve the measurement issue and estimate TFP growth.
2 September 2011
© 2011 Elsevier B.V. All rights reserved.
Accepted 22 September 2011
Available online 1 October 2011

JEL classification:
O11
O30
O41

Keywords:
Factor shares
Production function
Measurement

1. Introduction of substitution between factors is different from one; (iii) an


increase in the relative size of the sectors where factor shares are
The idea that labor income share is roughly constant, namely, different from the average or (iv) biased technological change.3 Of
the Cobb–Douglas–Kaldor paradigm,1 produced important conse- course, in the first case the functional distribution of income has no
quences in the area of economic growth. Almost all the literature relation with aggregate output so, if this is the correct explanation,
on growth accounting assumes that the elasticity of output with then there are no measurement problems. However, if any of the
respect to capital is constant and concludes that the major part of other explanations is correct then any change in the functional
economic growth is explained by growth in TFP (see Easterly and distribution of income is a result of a fundamental change that also
Levine, 2002; Solow, 1957 or Young, 1995 among others). affects aggregate output. 4
Empirical work shows that the Cobb–Douglas–Kaldor paradigm Changes in the factor shares have different effects on output
is not supported by the data. Kahn and Lim (1998) provide evidence depending on the factor abundance of the economy. If the income
that the shares of equipment, production workers and non- share of abundant factors is growing then the effect of this change
production workers have clear trends. Blanchard (1997) observes on the income level is positive but if the share of abundant factors
that the share of labor decreases in continental Europe after the decreases the effect of the change is negative.
80’s and suggests that the reason for such a decline may be To illustrate the importance of correct measures for the factors
technological bias. Other authors calculate the income share of consider the simplest Cobb–Douglas technology with two factors:
reproducible (human and physical capital) and not reproducible K and L. Output per worker, y = Y /L, can be expressed as a function
factors and argue that the latter is positively correlated with the of capital per worker k = K /L : y = Akα .
income level (see Krueger, 1999, Caselli and Feyrer, 2007, Zuleta, Now, suppose that there is an increase in the share of capital.
2008a, Sturgill, 2009 and Zuleta et al., 2010). The effect of the change in factor shares on income per worker
The variability of factor shares can be explained by different depends on the relative abundance of capital,
phenomena: (i) changes in the bargaining power of different ∂y
agents,2 (ii) an aggregate production function where the elasticity = Akα ln k.
∂α
∗ Tel.: +359 73 888 235.
E-mail address: hernando.zuleta@gmail.com. 3 In the last decades some authors revisited the theory of biased innovations and
1 See Cobb and Douglas (1928) and Kaldor (1961). challenged the Cobb–Douglas–Kaldor paradigm (see Zeira, 1998, 2005, Seater, 2005,
2 According to this line of research, the decrease in labor shares is due to a Peretto and Seater, 2006 and Zuleta, 2008b among others).
decrease in the bargaining power of workers generated by the institutional setting 4 According to Aghion and Howitt (2000) the attempt to attribute changes in
(see Bentolila and Saint-Paul, 2003, Giammarrioli et al., 2002, Berthold et al., 2002 human capital and raw labor shares to deunionization fails on the basis of time
and Bental and Demougin, 2010 among others). considerations.

0165-1765/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.econlet.2011.09.026
92 H. Zuleta / Economics Letters 114 (2012) 91–93

Table 1
A numerical exercise on the effect of changes in the capital share.
Capital stock per worker 1975 GDP per Change in Effect on income Percentage effect on GDP per
(constant prices 1985)a worker 1975 capital shareb worker 1975 (%)
(constant
prices 2000)a
Thousands (A) Ten (C) (D) Thousands Ten thousands Thousands Ten thousands
thousands (B) (ln(A) ∗ C ∗ D) (ln(B) ∗ C ∗ D) (E/C) (G) (F/C) (H)
(E) (F)

Australia 28.5 2.9 39 915.7 0.08 10 700.5 3347.7 27 8


Belgium 27.0 2.7 43 250.1 0.05 7 127.2 2147.8 16 5
Denmark 24.7 2.5 34 590.9 0.04 4 434.2 1248.3 13 4
Finland 28.9 2.9 32 293.3 0.10 10 865.2 3429.4 34 11
France 24.2 2.4 40 230.7 0.03 3 847.8 1068.7 10 3
Germany 36.9 3.7 38 315.2 0.05 6 912.5 2501.3 18 7
Ireland 13.9 1.4 30 533.0 0.17 13 676.4 1724.6 45 6
Italy 21.5 2.2 38 370.3 0.10 11 777.0 2941.9 31 8
Norway 38.9 3.9 46 667.9 0.11 18 797.6 6977.3 40 15
Sweden 25.6 2.6 37 406.9 0.02 2 426.2 703.6 6 2
United kingdom 14.6 1.5 33 511.3 0.10 8 988.6 1272.3 27 4
United states 26.1 2.6 48 774.3 0.02 3 182.3 936.2 7 2
a
Source: Penn world data tables.
b
Source: United Nations Data Set. The data was taken from Finnoff and Jayadev (2006). They used the United Nations dataset to construct the data on labor shares.

2. How to solve the problem?

2.1. Assuming away factor augmenting technological change

Consider a production function with only two factors, constant


returns to scale, decreasing marginal productivity of factors and all
the standard assumptions:
Yt = At F (φK Kt , φL Lt ) (1)
where Y is total income, K is physical capital, and L is raw labor.
Now, given that factor shares vary, the relative abundance of
factors becomes very important and, for this reason, it is necessary
Fig. 1. The effect of an increase in the capital share.
to have correct measures of capital and labor. The parameters φK
and φL play this role. If φK Kt ≥ φL Lt then the economy is capital
Therefore, if k > 1 then the effect is positive and if k < 1 the abundant.
effect is negative. Fig. 1 illustrates this fact. Differentiating Eq. (1),
In general, any change in the share of capital has an effect on 1Yt

1At 1Kt 1Lt

φK Kt

income. The size of this effect depends on both the capital labor = + αt + (1 − αt ) + 1αt ln (2)
Yt At Kt Lt φ L Lt
ratio and the income per worker. In Table 1 we provide a numerical
exercise that illustrates both the possible size of this effect and the where αt is the elasticity of output with respect to capital and
importance of the units used to measure the capital labor ratio for (1 − αt ) is the elasticity of output with respect to labor.
a set of OECD countries. The first two columns (A and B) present Now define
the capital per worker in 1975 measured in thousands and ten 1Yt 1 Kt 1Lt
 
thousands of dollars respectively (constant 1985 prices). The third St = − αt + (1 − αt ) . (3)
Yt Kt Lt
column (C) presents the GDP per worker in 1975 (constant 2005
prices). The fourth column (D) presents the change in capital share The variable S in Eq. (3) is the Solow residual. Now, from Eqs. (2)
between 1975 and 2000. Columns (E) and (F) show the effect that and (3) it follows that the Solow residual is not only TFP growth, it
the change in the capital share would have on the income level, also includes biased technological change:
1y = y ln k1α . In column (E) we compute the effect measuring 1At

φK Kt

the capital labor ratio in thousands of dollars and in column (F) we St = + 1αt ln . (4)
At φ L Lt
measure it in ten thousands of dollars. Finally, the last two columns
present these changes as a percentage of GDP per worker in 1975. Define S̃t = St − 1αt ln (Kt /Lt ), so Eq. (4) can be rewritten
The first thing to note is that when we measure the capital labor
1At φK
 
ratio in thousands the effect of the change in capital share is higher S̃t = + 1αt ln . (5)
than 10% of GDP per worker for 10 of the 12 countries. Second, At φL
when we measure the capital labor ratio in ten thousands, the size We can estimate the following equation:
of the effect is reduced substantially (for Ireland the difference is
close to 40%). Therefore, this numerical exercise suggests that (i) S̃t = C0 + C1 1αt + ρt (6)
the size of the effect of an increase in the capital share can be  
1At φK
significant in economic terms and (ii) the units we use to measure where At
= C0 + ρt and C1 = ln φL
.
the stock of capital affect in a significant way the estimation of the Therefore, this methodology allows us to identify the correct
effect. measures of factors per worker φK /φL and the growth rate of TFP.
H. Zuleta / Economics Letters 114 (2012) 91–93 93

Finally, from Eq. (1) we get: Note, however that in this case the ratio φK /φL may change
with time. In order to apply this methodology with constant units
φK Kt φK Kt
1−αt −αt
of measurement we need to choose a period t = v such that
 
Yt Yt
At φK = and At φL = . (7)
Kt φ L Lt Lt φ L Lt D1,v = D2,v = 1 and ln (φK /φL ) = C1 − ln (1 + C2 ).
In the same way we have to introduce the following restriction:
Therefore, knowing the correct measures of factors per worker,
C1 − ln (1 + C2 ) + ln 1 + C2 D2,t
 
we can also identify the level of TFP multiplied by the correct
measure of labor (and capital) for every t.
D1,t = .
C1

2.2. What if there is labor augmenting or capital augmenting 3. Conclusions


technological change?
The approach described above has three main advantages over
If there is factor augmenting technological change then Eq. (2) the standard growth accounting methodology. First, it takes into
becomes account the variability of factor shares. Second, it is also a way to
solve the problem of measurement. Finally, it helps to understand
1Yt 1At 1AKt 1ALt 1Kt

= + αt + (1 − αt ) + αt the possible effects that changes in factor shares have on aggregate
Yt At AKt ALt Kt output.
Applications of this methodology should take into account that
1Lt φK ,t Kt
 
+ (1 − αt ) + 1αt ln (2A) the correct units of measurement can be different from country to
Lt φL,t Lt country. Additionally, in the short run factor shares may fluctuate
1AKt 1ALt for reasons unrelated to the elasticity of output with respect to
where and are, respectively, capital and labor augment-
AK t ALt factors so the growth rates should be computed for time periods
ing technological change. of at least 5 years.
Eqs. (4) and (5) become
Acknowledgments
1At 1AKt 1ALt φK ,t Kt
  
St = + αt + (1 − αt ) + 1αt ln (4A)
At AKt ALt φL,t Lt I am grateful to Peter Howitt and one anonymous referee for
comments and suggestions. The usual caveat applies.
φK ,0 1AKt 1ALt 1At 1ALt
   
S̃t = 1αt ln 1+ − + + References
φL,0 AKt ALt At ALt
Aghion, P., Howitt, P., 2000. The Economics of Growth. MIT Press, Cambridge,
1AKt 1ALt
 
Massachusetts.
+ αt − . (5A) Bental, Benjamin, Demougin, Dominique, 2010. Declining labor shares and
AKt ALt bargaining power: an institutional explanation. Journal of Macroeconomics,
Finally, we can estimate the following equation: Elsevier 32 (1), 443–456.
Bentolila, S., Saint-Paul, G., 2003. Explaining movements in the labor share.
Contributions to Macroeconomics 3 (1), Article 9.
S̃t = C0 + C1 1αt + C2 αt + ρt (6A) Berthold, N., Fehn, R., Thode, E., 2002. Falling labor share and rising unemployment:
long–run consequences of institutional shocks? German Economic Review 3,
1AKt 1ALt 1AKt 1ALt
    
φK
where C1 = ln φL
1 + AKt
− ALt
, C2 = AKt
− ALt
and 431–459.
Blanchard, O.J., 1997. The medium run. In: Brooking Papers on Economic Activity,
1 At 1ALt Economic Studies Program. The Brookings Institution 28 (2), 89–158.
C0 + ρt = At
+ ALt
. Caselli, F., Feyrer, J., 2007. The marginal product of capital. Quarterly Journal of
Therefore, we can identify the correct measures of factors per Economics 122 (2), 535–568.
Cobb, C.W., Douglas, P.H., 1928. A theory of production. American Economic Review
worker φK /φL , the difference between capital augmenting and 18, 139–165. Supplement.
1AKt 1ALt Easterly, W., Levine, R., 2002. What have we learned from a decade of empirical
labor augmenting technological change AKt
− ALt
and the sum research on growth? it’s Not Factor Accumulation: Stylized Facts and Growth
1 At 1ALt Models. The World Bank Economic Review 15 (2), 177–219.
of neutral plus labor augmenting technological change At
+ ALt
. Finnoff, K., Jayadev, A., 2006. Feminization and the labor share of income. GEM-IWG
Working Paper 06-4.
Giammarrioli, N., Messina, J., Steinberger, T., Strozzi, C., 2002. European labor share
2.3. What if the rate of technological change is not constant? dynamics: an institutional perspective. EUI Working Paper ECO 13.
Kahn, J.A., Lim, J.S., 1998. Skilled labor-augmenting technical progress in U.S.
manufacturing. Staff Reports 47, Federal Reserve Bank of New York.
We consider four types of technological change: Change in Kaldor, N., 1961. Capital accumulation and economic growth. In: Lutz, F.A.,
Hague, D.C. (Eds.), The Theory of Capital. Martin’s Press, New York St.,
factor shares 1αt , neutral 1At , capital augmenting 1AK ,t and labor pp. 177–222.
augmenting 1AL,t . The change in factor shares is observed so there Krueger, A., 1999. Measuring labor’s share. American Economic Review 89 (2),
is no need to introduce changes in order to consider the possibility 45–51.
Peretto, P., Seater, J., 2006. Augmentation or elimination? Working Paper.
of changes in the growth of capital shares. Now, if factor shares http://ideas.repec.org/p/deg/conpap/c011_060.html.
are constant for the time period (and set of countries) under study Seater, J., 2005. Share-altering technical progress. In: Finley, L.A. (Ed.), Economic
Growth and Productivity. Nova Science Publishers, Hauppage, pp. 59–84.
then it is not possible to find the correct measure for the capital Solow, R., 1957. Technical change and the aggregate production function. The
labor ratio φK /φL . In other words, we need a sufficient number Review of Economics and Statistics 39, 312–320.
of observations where factor shares change in order to estimate Sturgill, B., 2009. Cross-country variation in factor shares and its implications for
development accounting. Working Papers 09-07. Department of Economics,
φK /φL . Appalachian State University.
Regarding the other types of technological change, it is possible Young, A., 1995. The tyranny of numbers: confronting the statistical realities of the
to use time dummies (D0,t , D1,t and D2,t ) and modify Eq. (6) in the east asian growth experience. Quarterly Journal of Economics 110, 641–680.
Zeira, J., 1998. Workers, machines and economic growth. Quarterly Journal of
following way: Economics 113 (4), 1091–1117.
Zeira, J., 2005. Machines as engines of growth, CEPR Discussion Papers 5429.
S̃t = C0 + D0,t + D1,t C1 1αt + D2,t C2 αt + ρt (6B) Zuleta, H., 2008a. An empirical note on factor shares. Journal of International Trade
& Economic Development 17 (3), 379–390.
1 AK t 1ALt Zuleta, H., 2008b. Factor saving innovations and factor income shares. Review of
  
φK
where C1 D1,t = ln φL
1+ AKt
− ALt
, Economic Dynamics 11 (4), 836–851.
Zuleta, H., Parada, J., García, A., Campo, J., 2010. Participación factorial y contabilidad
1 AK t 1ALt 1ALt
 
C2 D2,t = − AL and C0 + D0,t + ρt = 1AAt t +
AKt ALt
. del crecimiento económico en Colombia 1984–2005. Una Propuesta de
t Modificación del Método de Contabilidad del Crecimiento Desarrollo Y
Therefore, ln (φK /φL ) = C1 D1,t − ln (1 + C2 ). Sociedad, vol. 65, pp. 71–121.

You might also like