Economic Analysis and Policy
137
Vol. 16 No. 02, September 1986
A NOTE ON THE DETERMINANTS OF THE COST STRUCTURE OF GROSS DOMESTIC PRODUCT: A CROSS COUNTRY ANALYSIS M.M Metwally,
Department oj Economics, King Saud University. Riyadh,
Saudi Arabia, and
Rick Tamaschke, Department of Economics. University 0/ Queensland, SI Lucia, Qld. 4067. This note provides a cross section analysis to investigate reasons fo r the differences in the cost structure of Gross Domestic Product exhibited by some 45 countries with a view to shedding so me light on the determinants of these structures about which surprisingly little is known. The analysi s mak es use of some appropriate econometric models which encompass a number of theoretically plausible explanatory variables. Income per capita, population size, the share of the government sector and the degree of industralisation were found to be important determinants in "expla ining" differences in compensat ion of employees and operatin g surplus between countries. 1.
Introduction
There has been grow ing interest recently in the cost structure of Gross Domestic Product (GOP) by economists and policy makers alike . Surprising ly little is known about the determinants of cost structures and the whole questio n of investigating these determinants statisti cally has so far been sadly neglected.' In this respect two lines of inquiry arc of particular interest. The first is to trace the behaviour of cost structures, and factor shares, throug h time and to investigate the effects of variables suc h as inflation, technological change and monopoly power for individual countries using lime series data. Such work has appeared elsewhere. IM etwally and Tamaschke (1983») A seco nd approach is a cross section ana lys is using individual count ry data in a given yea r as th e unit s o f observatio n. This no te is directed at this second line of inquiry and provides a cross section study using the data of 45 co untries for which adequate statistical information is available. Thus this note complements the time se ries work of Metwally and Tamasch ke (1983) and as such, draws in part, on the data set employed in the earlier study, The analysis of this note begin s with Section 2 On the theory of distribution sec Fellner and Haley (1961), Hahn a nd Mathews (1968), Kaldor (1960) and Kalecki (1939).
Economic Analysis and Policy
138
Vol. 16 No. 02, September 1986
in which the main characteristics of the cost structures of the sample countries are examined . In Section 3 some econometric models are developed which examine th e impact of a number o f theo reti cally plausible macroeco nomic variables on cost structures. Fina lly Section 4 summarises the results of this study. 2.
Cost Structures in 4S Countries
Table 1 provides data on the cost structure of GOP for each of the 45 sample countries. Perusal of this Table shows that there are some substant ial differences between the relative shares. In group (I), which consists of Europe, North America, Japan , Australia and New Zea land , " compensation of employees" l is significant ly higher than "operating surplu s" with the exceptions of Greece and Malta which show on ly a comparatively small difference between the respecti ve shares. Emp loyees are " best off" in Norway, the United Kingdom and the United States. Group (2) , consist ing of Latin American and Caribbea n countries, generally shows the opposite characteristics to group (I), i.e. operating surplus exceeds compensation of employee ;, the main exceptions bei ng Nicaragua, Costa Rica and Pu erto Rico, but even here the data are well below the relative percentages exhibited by group (I). In group (3), Africa, Asia and other countries, the results are mixed. Thus in India, French Polynesia, South Africa and Za mbia, "labour's" sha re is clearly the greater whereas in some other countries, notably Sierra Leone, Solomon Island s and Thailand, the position is clearly the opposite. Looking at the groups as a whole, compensation of employees varies enormously between countries with a high of 69 per cent of GOP for India and a low of 25 per cen t in the case of Sierra Leone; operating surplus ranges from 57 per cent in the cases of Sierra Leone a nd Tanzania to 14 per cen t in the cases of Norway and India.
The UN stat LstLcs used for compensa tion o r employees purpOrt to inclu de a proportion of mixed Incomes and hence do not seem unreasonable est imates of labour income as required by economic theory. Thus this study could also shed some light on "factor shares". The data used are for 1977.
139
Economic Analysis and Policy
Vol. 16 No. 02, September 1986
TABLE 1 COST STRUCTURE OF GDP IN 45 COUNTRIES 1977
Country
(I)
Europe, No rth A merica, Japan, Australia & N ew Zealand Australia Austria Belgium Canada Denma rk Finland France Germany (W) Greece Ireland Italy Japan Malta Net herla nds New Zealand Norway Portugal Spain UK USA
Compensation of Employees as a % of GOP (A)
Operating Surplus as % of GO P
Relative Shares
(B)
(A) + (B)
56 54 58 57 58 55 55 56 36 55 57 54 47 59 55 59 59 57 62 62
26 20 23 21 22 24 22 22 45 27 23 27 41 22 29 14 28 31 15 17
.. 15 2.70
33 34 51 28 49 43 38 54 35 41 51 34 41
47 52 35 59 38 42 45 33 52 46 35 50 50
2.52
2.71 2.64 2.29 2.50 2.55 0.80 2.04 2.48 2.00 1. 15
2.68 1.90 4.21 2. 11 1. 84
4.13 3.65
(2) Latin America & Caribbean Chile Colombia Costa Rica Ecuador Guyana Hon du ras Mexico Nicaragua Paraguay Peru Puerto Rico Uruguay Venezuela
0.70 0.65 1.46
0.47 1.29 1.02 0.84 1.64 0.67 0.87 1.46 0.68 0.82
140
Economic Analysis and Policy
Vol. 16 No. 02, September 1986
TABLE 1 (continued)
Country
(3) Africa, Asia & Others Burma French Polynesia India Korea Papua New Guinea Sierra Leone Solomon Islands South Africa Tanzania T hailand Zambia
Compensation of Employees as a % of GOP
Operating Surplus as % of GOP
Relative Shares
(A)
(B)
(A) + (9)
39 52 69 34 44 25 26 53 28 26 48
45 21 14 46 44 57 58
0.85
27 57 55
29
2.48 4.93 0.74 1.00 0.44 0.45 1.96
0.49 0.47 1.66
Source: United Nations (1978) and Individual Countries' National Accounts Statistics.
Economic Analysis and Policy
141
3.
Vol. 16 No. 02, September 1986
So me Econometric Models
In an attempt to throw some light on the reasons for th e differences in cost structure displayed by the 45 countries included in Table I , the following statistica l models, which incorpora te a good number of th eoretically plau sible macroeconomic variables, were dcveloped.
The Models (I) Ai
=
(2) Bi
= ~O + ~ l yhi + ~ 2Ni + ~ lYi
(XO+Ctlyhj + Ct2Ni + (l3 Yi + Ct41i + (XSXi + Ct6G i C(7 ri + ((8Mj + Uj.
+ ~4 1 i + ~5X i
+ ~6Gi + ~7ri + ~ gM i + Vi· (l) (AlB)i
=
YO
+ Ylyhi + Y2Ni + YlYi + Y 4 1i + Y5Xi
+ 16Gi + 'Y7fi + 18Mi + Wi. where
A = compensat ion of employees as a percentage of GOP, B = operating surp lus as a pe rcentage of GOP, (A / B)= relative sha res, i.e . A l B, yh = GOP per head measured in US $s, = population in millio ns, N "" GOP in billions of US $s, Y I = gross fixed capita l fo rm ation as a percentage of GOP, X "" exports as a percentage of GOP, G = governme nt co nsumpti o n as a percentage of GOP, r = average rate of growth of r eal GO P , and M = manufacturing output as a percentage of GOP. i refers to the ith country. ui, Vi and wi are error terms.
Model ( I) focusses attention o n compensation of employees, model (2) on operati ng surplus a nd model (3) o n relative sha res. The eq uations were estimated by linear least squares methods. ) They were tested for multicollinearity a long the lines suggested in J ohnston (1984) and these tests gave no undue cause fo r concern . Th e residua ls about the fitted equations were a lso examined for heteroscedast icity and this analysis a lso gave no cause for concern." The various regression results are presented in Table 2. As the explanatory variables in equations (I) and (2) are identical, this approach is equivalent to the methods outlined in Zellner ( 1962). Besides inspection we also tried partitioning the sample residuals into different groups and testi ng for significant differences between the respective variances. On tests for hctcroscedasticity sec In{riligator (1978) and J ohnston (1984) . The Durbin Watson test allayed fears of {he possib le presence of first order serial correlation.
142
Economic Analysis and Policy
Vol. 16 No. 02, September 1986
TABLE 2
CROSS COUNTRY REGRESSION RESU LTS
A
=
11.185 (1.885)
n = 45;
B
=
+
0.0002 yh· (4.333)
+ 0.062 N·
+
0. 148 I (0.802)
+
0.287 r (1.328)
R2 '" 0.754;
(5 .346)
· 0.00001 Y (· 1.388)
+
0.007 X (0.093)
+ 0.700 G (3.576)
+
0.575 M· (3.649)
F'
.13.765.
0.000 1 yh· (·5.476)
4.628 (24.488)
+
0.005 I (·0.874) 0.006 r (· 1.565) n = 45; (AlB)
=
-0.746 (·1.382)
n :::: 45;
R2 = 0.799;
0.002 N· (·6. 126)
+ 0.0000 1 Y (1.326)
0.002 X (0.723)
· 0.026 G (·4.234) (2)
0.0 15 M· (·3 .055)
F'
= 17.829.
+
0,0002 yh· (4.9 14)
+
+
0.006 I (0.359)
+ 0.00 1 X
+
0.014 r (1. 212)
+ 0,025 M·
R2 = 0.765;
(I)
0,007 N· (7.092)
(0.176)
- 0.00004 Y (·1.083)
+ 0.055 G (3. 108) (3)
(2.313)
F' = 14.666.
No les : · Signiricant at least the 5 per cent level. Values in parentheses are " t" statistics.
143
Economic Analysis and Policy
Vol. 16 No. 02, September 1986
These results support the following views: \.
Compensation o f employess is positively re lated to income per capita. This. combined with the not significant GOP variab le, suggests that "labour's" share is close ly re lated to th e sta nda rd of living but is ind ependent of the sheer size of G OP . Operating surp lus is inversely related to income per capita.
2.
popu lat ion size is an importan t explanato ry variable in terms of cost structures across coun tries. The signs of the coefficients suggest that this variable has a positi ve effe ct on compensation of employees and a negative effect on opera ting surplus.
3.
Governm ent consumption, as a perce ntage of GOP. has a positi "e effect on " labour's" share. Thus, as on e might expect, the greater the weight of the pub lic sector. the greater the proportion of GOP going to compensat io n of employees.
4.
Manufacturing outpu t, as a percentage of GDP has a positive effect o n compensation of employees, a nd therefo re a negative effect o n operat ing surplus. This would seem to support the view that vary ing degrees of industralisation have a significant im pact o n the relative shares to the extent that labour in the more indu stralised countries receives a greater share of GDP.
5.
The rate of rea l GOP growth as well as exports and Gross Fixed Cap ita l Formation, both expressed as percentages of GOP, were fou nd to be not significa nt in terms of explaining the differences in cost structures betwee n count ries .
Overall the "explanatory" power of the equatio ns is pleasi ng. The "good ness of fit" statistics are not only high, bearing in mind th at the samp le size is 45, but also the signs and significance levels of individual coefficients are remarkably consiste nt across equat io ns, 4.
Summary
The purpose of this note has been to analyse the differences in the cost structures of GDP exhibited by some 45 cou ntries, for which the relevant data were available, with a view o r shedd ing so me lighl o n the determinants of these structures, about.which surprisingly little is known, The main result of this study may be summ ansed as follow s: 1.
There are so me major dirferences between the cost structures ex hibit ed by the 45 coun tri es st udied. Overall, co mpen sation of employees seems generally to be significantly higher in developed, co mpared with developing, economies.
Economic Analysis and Policy
144
2.
Vol. 16 No. 02, September 1986
Income per capita, population size ,. manufacturing output, as a percentage of GOP. and government consumptIOn , as a percentage of GOP, were found to be important in terms of explaining the differences in cost structures across countries. These variables are related positively to compensation of employees and negatively to operating surplus.
Overall the equations developed in this study seem to "explain " differences in cost stru ctures across countries quite successfully. However, it is to be hoped that this in vestigation will help encourage further research into what has so far been a sad ly neglected area of inquiry. REFERENCES Fellner, W . and Haley, B.F. (1961) (eds.) Readings in the Theory 0/ Income Distribution, (Allen and Unwin, London). Hahn, F.H. and Mathews, R.C.O. (1968) "The Theory of Eco nomic Growth: A Survey", in Surveys oj Economic Theory , Vol. II , (M acmillan, London) . Intril igator, M.D. (1978) Econometric Models, Techniques and Applications, (Prentice-Hall, Englewood Cliffs). Johnston , J. (1 984) Econometric Methods , 3rd Ed., (McGraw-Hili, New York). Kaldor, N. (1960) " Alternative Theories of Distribution", reprinted in Essays on Valu e and Distribution, (Duckworth, London). Kalecki, M. (1939) The Distribution oj National Income. Essays in the Theory oj Economic Fluctuations, (Allen and Unwin, London) . Metwally, M.M. and Tamaschkc, H .U. (1983) "The Effect of Inflation and Technology on Factor Shares", Applied Economics, Vol. 15 No.6, pp. 777-91. United Nations (1978) Yearbook oj National A ccounts Statistics, Vol. II, (United Nations, New York). Zellner. A. (1 962) "An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias". Journal 0/ the American Statistical Association, Vol. 57 No .4. pp. 348-68.