"The short-run behaviour of prices and quantities in Indian industry", in `Indian industry: policies and performance', edited by D. Mookherjee, Delhi: Oxford University Press, 1995.
In this paper I attempt to establish the determinants of aggregate price and output behaviour in the manufacturing sector of the Indian economy. It is an analysis of short-run behaviour in the sense that it abstracts from the question of growth and structural change. This, of course, depends upon the premise of a certain autonomy to short-run behaviour being true. I follow a data-based approach, in that the aim is to secure from the data an idea of industrial price and output behaviour in the Indian economy with a minimum of a priori restrictions.
A brief discussion of the theoretical issues with respect to pricing rules in industry appears in Section IIa. Some stylised facts of industry price-behaviour in India are presented in Section IIb and the results of an econometric exercise are presented in section IIc. In Section IIIa I discuss several influential accounts of output determination in Indian industry. A statistical evaluation of these follows in Section IIIb. The conclusion in section IV contains a discussion of the macroeconomic implications of the price and output determination mechanisms that characterise the Indian manufacturing sector.
II. Industrial price behaviour
IIa. Theory
A great deal has been written about the theory of industrial pricing. I shall only be concerned with those aspects of this literature that has a bearing on the questions I address in the
empirical investigation to follow.
I take it as axiomatic that price-taking is not a credible description of the pricing rule in industry. In fact, some version of price setting would best describe industrial product markets. Practitioners have often taken the view that industry is an area of the economy characterised by `mark-up' pricing, i.e., that here firms set prices by adding a percentage mark-up to some clearly defined measure of costs. It is not immediately clear as to what this must imply. For the mark-up is the excess of price over marginal cost that emerges under non-competitive market structures. Would this then imply that all pricing outside perfect competition is mark-up pricing? I shall argue that this is not so, and that the essential question relates to the determination of the mark-up.
A closer reading suggests that the proponents of mark-up pricing are also suggesting something about the objective function of firms. This may be surmised from the proposition that the pricing rule, which invariably boils down to the decision regarding the percentage mark-up, is invariant with respect to demand conditions. It is implicit in the idea of a cost-determined price. This needs to be emphasised, for, the use of cost functions, a part of the tool kit of standard economic theory, itself would allow for a change in the price following upon a change in costs. From such a perspective, therefore, the `theory' of mark-up pricing cannot be seen to make any distinctive point at all.
In the writings of Kalecki it was explicitly assumed that, owing to uncertainty, firms do not attempt to maximise profits. Assuming further that cost curves are inverted L-shaped, that is, the marginal cost is constant over the relevant range of production, Kalecki argued that the price-cost relation was unaffected by short-run variations in demand. Excess capacity and constant marginal cost enabled firms to expand output and maintain price as demand varied. A departure from stable mark-ups was envisaged only under a situation of an extreme variation in demand. In the boom, the emergence of "cut throat" competition was expected to lead to a decline in the mark-up. In the recession, Kalecki surmised, that firms might adopt tacit collusion to not reduce prices to the same extent as the reduction in costs with a view to protecting their net-profit margins. If overheads, which remain constant over the cycle, are not included in `prime costs' a rise in the price-cost relation is needed to protect profitability. Since firms are assumed to mark-up prime costs, net margins are maintained during the recession only if aggregate proceeds are allowed to increase, which is possible only with an increase in the price-cost ratio. Note that both these are instances of a countercyclical mark-up. Kalecki was however aware of these being extreme instances with regard to the specification of demand conditions, and should be read as suggesting that there is a tendency for mark-ups to vary under the circumstances considered.
The question of the determination of the mark-up was however left somewhat open in Kalecki's writing. As Kaldor has suggested there is a circularity involved in arguing that the price-cost relation (the mark-up) is determined by the degree of monopoly while defining the degree of monopoly as [(p-c)/p] which is precisely the Lerner expression for the mark-up. Nevertheless, Kalecki's work had provided some original insights into the behaviour of firms with respect to pricing.
Of course, more conventionally, the question of the behaviour of the mark-up has also been investigated within the framework of profit maximisation. For instance, consider the problem of a Cournot oligopolist producing a homogenous product:
where p is the price of the product, X is industry output and x_{i}
is the ith firm's output.
The first-order condition for a maximum:
where c_{i} is the marginal cost of the ith firm, can be re-written to yield the basic Cournot-oligopoly pricing formula:
Now in the case where all firms have identical cost functions (3)
becomes
where ç is the number of firms and å is the elasticity of demand as before. When cost functions are no longer identical across firms the industry-wide average mark-up is just the weighted average of the individual firm mark-ups, the weights being the market shares:
Substituting for firm i's mark-up from (3), we have
where H is the Herfindahl index of concentration. Note how in the above representation market structure is captured by the two parameters H and å. What emerges very clearly is that the mark-up is inversely related to the elasticity of demand and the number of firms. It is directly related to the degree of concentration in an industry. From (4) we see that as the number of firms tends to one the formula for the monopolist is replicated while as the number of firms tends to infinity the mark-up tends to zero, as it must under perfect competition. This is in conformity with intuition on this question. The above discussion must set at rest the presumption that there is something less than rational about firms maintaining price in the face of demand shifts, even though it is demonstrated only for Cournot, rather than collusive, behaviour and assumes an unchanging elasticity of demand over the cycle. Under profit-maximising behaviour the consequence of a change in demand (i.e., a shift in the demand curve) is entirely governed by the contemporaneous change (if any) in the elasticity of demand. The early view was that this elasticity varied inversely over the (trade) cycle. This had led Harrod to predict pro-cyclicality of mark-ups. More recently, however, arguments have been produced to predict a pattern quite the reverse of this. That is the elasticity is believed to rise in the upswing because unattached customers search for new products. This would render the mark-up countercyclical.
Strategic behaviour on the part of firms has been the focus of more recent work on the behaviour of the mark-up over the cycle. Two accounts may be noted. Both these predict counter-cyclical movement of the mark-up. An argument by Stiglitz uses the familiar framework of `entry-limit' pricing. Consider an instance of recession. Now the existence of excess capacity reduces the threat of entry, and the price-cost ratio may increase relative to the boom when capacity utilisation is nearer to the limit. Another argument, by Rotemberg and Saloner is cast in terms of the problem of policing collusive arrangements. The hypothesis is that the ratio of the actual to the monopoly price is sensitive to the cycle. It is argued that in the boom this ratio is likely to be lower, i.e., we have `competitive' outcomes, for the firms have an incentive to undercut the industry price. Booms represent periods when firms' incentive to deviate from the collusive outcome (of maintaining price) is greater, for the punishment must be felt in periods with lower demand, and thus lower profits. Now, given costs, a counter-cyclical behaviour of the actual to the monopoly price implies a counter-cyclical price-cost ratio. Rotemberg and Saloner present evidence of the latter from the U.S. economy. Even more interestingly, they also provide evidence that suggests that price wars are more likely to occur during the boom. Note that the idea is close to that of Kalecki discussed earlier on in this essay.
This brief survey was intended to convey that economic theory does not provide an unambiguous answer of the behaviour of the mark-up over the cycle. We have a situation ripe for empirical investigation.
IIb. Prices, costs and activity: the stylised facts
In Chart 1 are presented the contemporaneous variation in the indices of industrial prices output. Apart from a single phase, namely the period 1957-62, there is a striking inverse relation between changes in the two variables, suggesting a countercyclical mark-up. While this information is far from adequate to conclude on the behaviour of the mark-up over the cycle it does point to an aspect of the relation between growth and inflation that is often ignored. In the on-going debate on the appropriate stabilisation policy for the Indian economy it is all too easily assumed that putting a brake on growth is a way out of inflation. I shall return to a consideration of this theme later on in this essay.
[CHART 1 GOES ABOUT HERE]
In Chart 2 are presented the rates of change in price and the two major components of prime costs, namely, labour and materials cost, respectively. Note that price and costs move together. However, the change in price is clearly a little more stable than the changes in costs. Note that `materials' has been defined as `raw' materials including oil (see Appendix for details). This must be borne in mind when studying the graphs, for it will be noted that the change in prices does not track the change in materials cost so closely as it does the change in labour costs. This is partly explained by the fact that the expenditure on wages constitutes a much larger share of the outlay on costs than the expenditure on raw materials as opposed to materials (including intermediate inputs) as a whole. It also points to a feature of the economy that I have stressed, i.e., that certain prices are more responsive to excess demands than are others. (Of course, that raw-materials prices are driven by excess demand is not something that can be inferred from the chart. I have, however, elsewhere provided evidence of an econometric nature that this is so. The relative stability of labour costs is due to the link between wages and the cost-of-living which includes the industrial price. The instability in labour costs is obviously due to the variation in productivity, for which no explanation is ventured in this study.
[CHART 2 goes about here]
IIc. Price behaviour in Indian industry: an econometric
investigation
In this section are presented estimates of price equations for Indian industry. The econometric methodology adopted is the well-known Engle-Granger `two step' method associated with the idea of co-integration. I start with a brief comment on the concepts and the methods.
While the idea of co-integration is essentially statistical in its origins it has been imaginatively exploited in the study of the behaviour of economic variables over time. I discuss each of these aspects.
A large part of the variables that economists are likely to be interested in are non-stationary, in the sense that they possess time dependent unconditional mean and variance. These integrated variables require to be appropriately de-trended before being rendered stationary. In the presence of stochastic trends this is attained by differencing. This motivates the notion of the `order' of integration: the time series of a variable y_{t} is said to be integrated of order d [or, y » I(d)] if its time series attains stationarity after being differenced d times. Thus, for example, a time-series integrated of order zero, denoted I(0), is stationary in levels while for a time-series integrated of order one, denoted I(1), it is the first difference that is stationary. The idea of `co-integration' is now easily exposited: consider two time series, y_{t} and x_{t}, which are both I(d). Being of the same order of integration they possess the same long-run properties. Now, if there exists a vector (1, - â)' such that the combination
z_{t} = (y_{t} - âx_{t}) is I(d-b), b > 0
then y_{t} and x_{t} are said to be co-integrated.
The concept of co-integration may be seen to mimic the existence of a long-run equilibrium to which the system converges over time. To appreciate this, assume that economic theory suggests the following relationship between the variablesy_{t} and x_{t}:
Now z_{t}, which has been defined above, can be interpreted as the `equilibrium error', or, the distance the system is away from equilibrium.
In an important paper, with wide-ranging implications for applied economic research, Engle and Granger have shown that if two variables y_{t} and x_{t} are co-integrated then the relation between them may be represented in the `error-correction' model form:
Note that z_{t-1}, the lagged equilibrium-error, is one of the
determinants of the dynamics of the system. This is the idea of error correction, i.e., the dynamics correct for any inherited
disequilibrium, which is the `equilibrium error'. The isomorphism between `co-integration' and `error correction' is now easy to see. In terms of the current example, co-integration would imply that while x_{t} and y_{t} are individually non-stationary, and of the same order of integration, z_{t} is itself stationary (assuming that x_{t} and y_{t} are themselves I(1)). Stationarity of the z_{t} series means that deviations from a putative equilibrium are stationary, i.e., that there is error correction, or that the two variables move together in the long run. This conveys the essence of the Engle and Granger Representation Theorem.
Engle and Granger have recommended a procedure of estimation of dynamic equations that has come to be known as the `two-step method'. The first step is to run an OLS regression of the relationship postulated, ideally one suggested by economic theory. This is very likely to be a equation in the levels of the variables, rather like (7) above. This is referred to as the co-integrating regression. The estimated residuals of this regression are now tested for stationarity. If found to be stationary, the variables included in the regression are said to be co-integrated. The next step in the two-stage procedure would be to now estimate the equivalent of (8) with the lagged values of the residuals from the co-integrating regression included. In the history of the development of econometric methodology this procedure was also seen as a `reply' to the very powerful Box-Jenkins methodology then prevailing, in the sense that the ECM, it was argued, forecast better. But the procedure is also to be seen as a crucial element in the testing of the predictions of Economic Theory. This is the role of error-correction terms in the dynamic specification of econometric models.
A necessary step in the modelling of economic variables as co-integrated systems is that of establishing their time-series properties, notably the order of integration of each. This is done here by performing the unit-root tests under the Dickey-Fuller procedure, and the results are presented in Table 1. The augmented Dickey-Fuller test was conducted but the lagged-dependent variable was found to be statistically not significant mostly. Note that the tests suggest that all variables, other than the activity terms are stationary in their first differences, ie., they are I(1). The activity terms are stationary in levels, which, they being capacity utilisation terms, is exactly as we would expect. All variables have been entered in logarithms.
Table 1
Time series properties of prices, costs and activity
variable\null |
I(0) |
I(1) |
p |
-2.97 |
-4.86 |
l |
-2.08 |
-4.52 |
rm |
-2.90 |
-5.82 |
p_{f} |
-3.33 |
-4.71 |
q |
-2.46 |
-8.75 |
d |
-5.18 |
- |
y |
-2.42 |
- |
Dickey-Fuller statistic; `I(0)' and `I(1)' indicate that the null is that the series is non-stationary in levels and in 1st
differences, respectively; the regression is a first-order UNI-
VAR: regression for q excluded trend, regression for d exc-
luded constant and trend; T = 31; 5% critical value(s) of D-F
test statistic for T = 25 is -3.60 for full regression, -3.00
with trend excluded and -1.95 for constant and trend excluded.
See Fuller (1976).
I now proceed to the co-integrating regression. Two alternatives were run and the associated statistics are presented in Table 2. First, notice that these regressions differ according to the treatment of labour costs. While the first enters this variable (l) by itself, the second disaggregates labour costs into wages (w) and productivity (q). Note that, since all variables are entered in logarithms, l = (w - q). We can say little going by the co-integrating regression Durbin-Watson statistic (`CRDW') because it falls in the inconclusive region for both regressions. On the other hand, the value of the Augmented Dickey-Fuller statistic implies rejection of the null of non-stationarity of the residuals in the co-integrating regressions, in turn validating the positing of an equilibrium relationship between price and costs. But since the critical values are available only for the case of 50 observations and above, while the present exercise is restricted to 31 observations the verdict is given with greater confidence in the case of regression 2. However, given the widely acknowledged low power of the extant tests for co-integration, I shall work with the residuals of both the regressions reported in Table 2.
Table 2: The co-integrating regression
Regression |
CRDW |
ADF |
I |
.89 |
-3.82 |
II |
1.37 |
-5.62 |
I: p = a_{0} + a_{1}l + a_{2}rm + u, II: p = b_{0} + b_{1}w + b_{2}q + b_{3}rm + v;
`CRDW' is the co-integrating regression Durbin-Watson statistic and `CRADF' is the co-integrating regression augmented Dickey-
Fuller statistic; T=31; critical values at 5-percent level:
CRDW (n=3, T=31) bounds are .70 and 1.74, CRADF (n=2 and T=50)
= 3.67, and CRADF (n=3 and T=50) = 4.11. [Engle & Yoo (1987).]
Before proceeding to the estimates of the dynamic specification of the price equation, i.e., the second stage of the two-stage procedure discussed above, some considerations that prevailed upon the choice and construction of the variables are discussed:
(a) Undoubtedly the most important question is the measurement of demand. Orders on firms' books would be the obvious way to measure demand, but this information is not available. In studies on price behaviour in industrialised economies, some output-based measure is used as a second-best procedure. In the Indian context, however, this may not be without problems for it has been suggested that here industrial capacity-utilisation is often supply constrained. To that extent, output-based measures are not representative of demand conditions. So two measures of demand are to be used here. One is the deviation of national income from its trend value, denoted y and the other is the deviation of industrial production from its trend value, denoted d. Clearly the former would be the more appropriate measure of the state of demand faced by the industrial sector if production is supply constrained.
(b) No special allowance is made for international demand per se (apart from the extent to which it can be captured by the capacity utilisation index) or for the role of international competition via international prices. It seemed reasonable to take the view that for the period studied here, itself a matter for discussion later on in this paper, neither export demand nor international prices formed a significant element in the determination of the demand for Indian industrial production.
(c) As for the modelling of costs, while labour cost is measured as wages adjusted for productivity, the cost of material inputs is represented by an index of the price of raw materials. While this reflects the widespread practice of assuming that there is no variation in the efficiency of materials input use, it may not be entirely appropriate. The use of a `raw' materials index itself is more easily justified. Using the 'materials bill per unit output' in a regression would have meant explaining industrial prices partly by itself. A `raw materials' index thus measures the true materials input cost of industry as a whole. While on this question, it may be added that the index constructed here includes the price of oil and other raw material imports. Thus, even though foreign demand is excluded from the specification, because it is not considered to be a major factor over the period studied here, the relevant aspects of the external sector are not ignored.
Estimates of price equations for aggregate industry are presented in Table 3. First, the econometric considerations. An explanation is in order for the inclusion of the rate of change of activity in all equations. Strictly, both considerations based on the time-series properties of the variables and those based on theoretical reasoning would have suggested restricting ourselves to the lagged level of activity. For since the level of activity is I(0), as one would expect of the deviation of output from its trend value, it is a perfect match for the dependent variable, the rate of change of price, also I(0), implying that there is no need for any further differencing of the activity term. Moreover, recall from the earlier part of this essay that the theoretical formulations all relate the mark-up to the level activity, as it is measured by the degree of capacity utilisation. The reason for the retention of the rate of change of activity must therefore be rationalised on statistical grounds. The estimate I.a worsened considerably when
Table 3
A price equation for Indian manufacturing industry
Var.\Eqn. |
I.a |
I.b |
II.a |
II.b |
Äl_{t} |
.57 (3.89) |
.66 (4.29) |
.52 (3.35) |
.67 (4.14) |
Ärm_{t} |
.37 (4.74) |
.29 (3.65) |
.40 (4.94) |
.29 (3.58) |
z_{t-1} |
-.37 (2.69) |
-.46 (2.34) |
-.47 (3.27) |
-.56 (2.81) |
d_{t-1} |
-.24 (2.99) |
-.14 (1.64) |
-.35 (2.51) |
-.11 (0.88) |
Äd_{t} |
-.33 (2.00) |
-.26 (1.39) |
-.23 (1.46) |
-.06 (0.36) |
D-W |
2.00 |
1.89 |
1.77 |
1.64 |
s.e.e. |
.0307 |
.0327 |
.0313 |
.0341 |
T = 1952-82, n = 31, `IV' estimates; t-values in parentheses; z in equations `a' is the residual from co-integrating regre-
ssion 1 in Table I, while z in equations `b' is the residual
from co-integrating regression II in Table 2; d in equations
`I' is the deviation of industrial production from trend, while
d in equations `II' is the deviation of national income from
its trend value.
this variable was dropped. Hence it was retained in all specifications to engender comparability. I now turn to the question of the preferred specification. Recall that specifications `I' and `II' difer according to the definition of their respective activity terms. So the choice is not among these. It is between estimates `a' and `b'. Recall that these differ according to the origin of the error-correction terms z. I restrict the criteria of choice among these to the residual statistics, and find that estimates `a' are clearly superior. Recall that the z in the `a' equations are residuals from the co-integrating regression 1 in Table 2. Not only is the specification of that regression intuitively more appealing than the specification of regression 2, but also the low power of the standard tests for co-integration suggests the desirability of supplementing these tests with information drawn from the estimate of the dynamic specification. On these grounds I choose estimates `a'. It may be added that estimate I.a has particularly desirable properties. The cost coefficients sum to very close to one, the residual statistics (D-W) signal zero auto-correlation, and a split-sample test (`Chow') for parameter stability turned in a statistically insignificant value [F (5, 21) = 2.68].
An interpretation of the results is straightforward. In the light of the concerns of this essay, I focus upon one aspect of the estimates, namely, the behaviour of the mark-up over the cycle. Notice right away that the activity terms are consistently negative though not always statistically significant. There is no evidence whatsoever for the belief that the aggregate industry mark-up responds directly to demand conditions. In fact, it is possible to conclude unambiguously, on the basis of estimate I.a, that it is countercyclical. However, for claims of alternative mark-up behaviour in Indian industry see Chatterji (1989) and Madhur and Roy (1986).
It is often suggested that the growth of the money stock is a determinant of industrial price behaviour. This proposition has dubious theoretical foundations, for the Quantity Theory of Money
is at best interpreted as saying that the price level is proportional to the money stock. It leaves no room for the inclusion of money in equations for sectoral price equations, which we habitually specify as price relatives. Inclusion of money in the industrial-price equation, as specified here, suggests that money affects the mark-up. There is no such suggestion in Monetary Theory. However, popular notions are seldom stymied by disclaimers on the grounds of economic theory! I have therefore re-estimated the specification I.a with the rate of growth (current and lagged) of the money stock replacing the demand terms. Two measures of the money stock (M_{1} and M_{3}) were used in alternate runs. The results are presented in Table 3.1.
Note that the money supply terms are not statistically signi-ficant. If `the forces of competition' are to be interpreted as making for price to adjust in the direction of excess demand, then they are clearly rather weak in the case of Indian industry.
Table 3.1
Money and prices
Variable |
I |
II |
Äl_{t} |
0.54 (3.46) |
0.60 (3.40) |
Ärm_{t} |
0.36 (4.61) |
0.36 (4.47) |
z_{t-1} |
-0.31 (2.23) |
-0.36 (2.55) |
Äm_{t} |
0.21 (0.96) |
0.08 (0.37) |
Äm_{t-1} |
-0.16 (0.62) |
-0.01 (0.03) |
T: 1952-80, N= 28, Dependent variable is Äp_{t}_{.} All variables are
in logarithms. Regressions `I' and `II' use the M_{1} and M_{3}
definitions of the money stock m, respectively.
III. The behaviour of industrial output
IIIa. Explanations of industrial growth in India:
There is a long tradition in Indian Economics of analysing industrial activity within the framework of two sector models.
Just as the dual-economy model of Lewis had at one stage provided a starting point for the analysis of growth in a developing economy, the more recent two-sector models are, no doubt, useful from the point of view of the empirical analysis of its short-run behaviour. Perhaps the most acclaimed is the one by Kaldor. The structure of the model is simple, revolving around a dichotomy between the agricultural and industrial sectors. There is a stylised asymmetry between these with respect to their price and output determination. Agriculture's output is supply determined while its price is demand determined. Exactly the opposite prevails in industry, with demand-determined output and supply conditions (costs) determining prices. With exogenously given (read `supply-determined') agricultural output, the model can be solved for the equilibrium values of industrial output and the relative price of the economy. Such models can even be extended, by the addition of a wage equation, to account for inflation. However, while the original Kaldor model remains an ingenious device for analysing short-run variations in industrial output, it is incomplete in its coverage of the factors that influence this output. In general, it shares with all sectoral models the characteristic that it ignores the role of macroeconomic variables in the determination of sectoral prices and output. In discussions of the Indian case it has been rectified by an emphasis on public investment as a demand-generating factor. The experience of the Indian economy in the eighties suggests that actually public investment might be too narrow a variable to concentrate upon when we are concerned with the demand-generating role of government. For, arguably, it is the government budget-deficit that we should be looking at. In India in the eighties we witnessed billowing budget deficits and the fastest rate of growth of industry since the 1950s. This would be entirely in keeping with the predictions of a Keynesian model. Of course, a different perspective on the role of government with respect to industrial growth in India has also been provided. Focusing on investment in infrastructure, especially in the key sectors of power and transport, Ahluwalia has argued that this eases a chronic supply-constraint in the Indian economy. The evidence presented by Ahluwalia does point to greater output in these sectors in the eighties, though the task of yet establishing the binding constraint on growth would remain.
[Chart 3 goes about here]
Chart 3 presents annual rates of change of public investment and industrial production. Notice that while a positive relation can be discerned, the swings in public investment are not matched by extent of variation in industrial growth. Another way of looking at this would be to say that industrial growth is being steadied by other factors, and it is to this question that I now turn.
In an economy with the structure of India it may reasonably be expected that the single most important determinant of the variation in industrial production must be the performance of agriculture. This is easy to see. First, agriculture contributes some crucial raw materials to industry. This would, however, be a declining influence as industrial production becomes progressively less dependent on raw-material inputs. After all, agriculture contributes the largest, even though declining, share of the national income. From the demand side it affects industrial production in two separate but mutually re-inforcing ways. Fluctuations in agricultural income alter the surplus available for expenditure on non-agricultural goods. After all, agriculture contributes the largest, even though declining, share of the national income. These fluctuations set in motion price adjustments which could have an effect on industrial output. In fact, `the terms of trade' argument relies on precisely such a mechanism. Chakravarty and Mitra have both argued that shifts in the inter-sectoral terms of trade against Industry affect industrial production adversely. Interestingly, both emphasise the effect on Industry's supply side. Chakravarty has argued that adverse terms-of-trade impinge upon Industry's savings while Mitra speaks of `industrial atrophy' due to declining profits. However, it is also possible to imagine a mechanism whereby agricultural prices affect industrial output from the demand side.
When not all incomes in the economy are indexed, changes in agricultural prices could give rise to income effects. These take the form of altering expenditure on industrial commodities. Essentially the rise in the price of food products must crowd out other expenditure from the budgets of fixed-income households. Thus partial equilibrium analyses suggest that, for instance, a dip in agricultural production would lower industrial production via two routes. The agricultural surplus is lower and economy-wide income effects, due to a rise in agricultural prices, reduce the demand for industrial goods. However, in general equilibrium it is conceivable that the demand for industrial products may rise after a decline in agricultural production. This is so because the value of the agricultural surplus may now be larger (following from the rise in price), and a higher demand from agricultural surplus households might more than compensate for the lower demand of fixed income households. This, of course, ignores the question of the number of households involved in the re-distribution.
It is easy to see that the shrinking demand efect is not really due to shifts in the terms of trade per se but due to inflation. Of course, since most periods of inflation are also periods when the terms of trade shift in favour of agriculture there is an observational equivalence, as it were.
The question of the impact of shifts in the internal terms-of-trade, essentially the relative price in a two-sector model, has been given much importance by researchers on the Indian economy, and it would be of interest to see the econometric evidence on this. Recent developments in econometric metho-dology have made us more sensitive to the time series properties of the variables we study, and this has often aided in the assessment of explanations. The present case is one such. For we see, from unit-root tests not reported here, that the terms of trade is a stationary variable while industrial production is a non-stationary one. This suggests that the terms of trade cannot be a determinant of industrial production in equilibrium. It can only induce fluctuations in industrial production. This vital distinction has not been sufficiently clearly appreciated, perhaps due to the considerable prestige accorded to arguments invoking the terms of trade.
IIIb. The determination of industrial output: an econometric
investigation
The approach taken here to the econometric investigation of output behaviour is a good deal less formal than it was in the case of price behaviour. This approach is, in fact, the conventional one of checking the data for the validity of some of the arguments put forward as explanations of the behaviour of industrial output in India. These arguments I have surveyed in the preceding section. Three variables emerge as being of interest as likely influences on industrial production in the short-run, viz., agricultural production, public investment and the terms of trade.
Denoting the level of industrial production by X_{i}, the level of agricultural production by X_{a} and the level of public investment by G, the following long-run relationship is postulated:
Note that the terms-of-trade have been excluded, since we are interested in the equilibrium relationship, and, as mentioned earlier, the terms-of-trade are a non-stationary variable. Co-integration is not rejected. Therefore, a dynamic specification in ECM-form was adopted. `General to specific' modelling yielded the parsimonious representation presented in Table 4. Since the price term enters with a lag the only likely source of simultaneity is eliminated and ordinary least-squares estimation was considered sufficient on this score. An explanation regarding the inclusion of the terms of trade in both levels and in change form is necessary. Being non-stationary, the inclusion of the lagged level would have been considered sufficient. However, letting remain the change form of the variable was necessary to yield statistically
Table 4
The determinants of industrial output
ÄG_{t} |
.08 (2.34) |
ÄXa_{t-1} |
.15 (2.44) |
Äè_{t-1} |
.24 (2.49) |
ecm |
-.12 (2.86) |
è_{t-1} |
-.01 (8.14) |
R^{2} |
.87 |
T = 1952-80, n = 28; OLS; error-autocor-
relation: Chi-square (1)=.213,
s.e.e. = 0.026; all variables are in
logarithms; `ecm' is lagged residual of
the relationship postulated in (9);
t-values in parentheses; parameter cons-
tancy (1967-80): Chow F(13, 10) = 0.56.
significant coefficients on `agricultural production' and `public investment'.
The set of explanations discussed in section IIIa seems to hold up. First, there is the effect on industrial output of public investment. However, the specification does not allow us to separate its demand-generating role emphasized here from its possible role in easing the supply constraint. When, to allow for a wider role for government, the equation for industrial production was re-estimated having replaced `public investment' by the `budget deficit' the latter was found to be not statistically significant.
Agriculture apparently affects industrial production in both the ways that have been imagined, i.e., both the agricultural price and output are important. Since the price term used here is the terms-of-trade of the Manufacturing sector, i.e., the ratio of the prices of manufactures to that of agricultural goods, the estimates imply that a shift in the
terms-of-trade in favour of agriculture would reduce industrial output. This is exactly as predicted in the writing on the subject discussed earlier on. Note that (the change in) both agricultural output and the terms-of-trade enter the equation with a lag. While this may be explained away in the case of the former as an artifact resulting from the fact that the agricultural year straddles two industrial production (i.e., financial) years, it is difficult to rationalise the time pattern with respect to the price term. A large part of the agricultural crop of one year might well reach the market only in the following financial year. However, the lagged effect of the change in the relative price suggests that agents take about a year to adjust their expenditures. This would appear unduly long. However, there would seem to be no reason to doubt the sign of the co-efficient on this term, for the lagged level of the same is negative too.
Statistically, we appear to have quite a satisfactory equation. However, it ought to be seen as being concerned with the question, in conventional terms, of what shifts the demand curve. Whether the binding constraint on Indian industry is a demand or a supply constraint is not dealt with here. Nor are we in a position to decipher whether the observed effect of the terms of trade is a pure relative-price effect or, as I have made out to be possible earlier on, an inflation effect.
IV. Conclusion: The macroeconomic implications of industrial
price and output behaviour
The mechanisms that govern industrial price and output determination in turn have implications for macroeconomic outcomes, and thus for the efficacy of alternative macro policies. The estimates presented here may be used to conceive of likely scenarios for India.
First, we can say that positive agricultural supply shocks increase economy-wide output and lower the inflation rate. By similar reasoning, negative shocks are stagflationary in their impact. We see from the output equation that a slower growth of agricultural production lowers the industrial growth rate. The accompanying rise in agricultural prices imply higher costs to industry. Estimates of the industrial price equation suggest that this will be passed on, regardless of demand conditions in the industrial markets. In fact, there might be some worsening of the situation, with firms actually raising mark-ups during the recession.
Secondly, at least as far industrial prices are concerned, the prospects for demand-based anti-inflationary policy must revolve around the extent to which it can affect the reduction of costs. This is so because the industrial mark-up is countercyclical and, as the evidence shows, money-wage determination in the organised sector is not affected by the state of activity. Lower activity levels do influence a crucial element of costs though, namely, that of raw materials. Output reduction, therefore, appears as the only route whereby demand-based strategies can counter inflation in the industrial sector. Price and output determination mechanisms in the industrial sector provide no scope to scale-down prices leaving quantities unchanged, whether we consider purely macroeconomic strategies such as contractionary fiscal policy or a sector-specific strategy such as targetted monetary policy, in particular, attempts to
ration credit. Essentially, demand-based strategies are weak in the face of rising industrial prices which can only be curbed by applying the brake on rising costs.
From a longer-run perspective, the containment of labour costs via continuous increases in labour productivity is a distinct possibility. This issue has been given little importance in the Indian context. For, ignoring raw-material costs, it is easy to see that where firms are unwilling to take a cut in their profit margins and unions are unwilling to acquiesce in real-wage reductions, the only prospect for pegging the industrial price is via productivity growth.
-------------
References
Ahluwalia, I.J. (1985) `Industrial growth in India: stagnation since the mid-sixties', Delhi: Oxford University Press.
Ahluwalia, I.J. (1991) `Productivity and growth in Indian manufacturing', Delhi: Oxford University Press.
Arrow, K.J. (1959) "Towards a theory of price adjustment", in M. Abramovitz (ed.) `The allocation of resources', Stanford: Stanford University Press.
Balakrishnan, P. (1991) `Pricing and inflation in India',Delhi: Oxford University Press.
Balakrishnan, P. (1992) `Monetary policy, inflation and activity', Bombay: Reserve Bank of India.
Chakravarty, S. (1973) "Reflections on the growth process in the Indian economy", in C.D. Wadhva (ed.): `Some problems of India's economic policy'. New Delhi: Tata McGraw Hill.
Chatterji, R. (1989) `The behaviour of industrial prices in India', Delhi: Oxford University Press.
Engle, R. and C. Granger (1987) "Co-integration and error-correction: representation, estimation, and testing", `Econometrica', 55, 251-76.
Engle, R. and C. Granger, ed., (1991) `Long run economic rela-tionships', Oxford: Oxford University Press.
Engle, R. and S. Yoo (1987) "Forecasting and testing in co-integrated systems", `Journal of Econometrics', 35, 143-159.
Fuller, W.A. (1976) `Introduction to statistical time series',
New York: Wiley.
Harrod, R. (1936) "Imperfect competition and the trade cycle", `Review of Economics and Statistics', 18, 84-8.
Kaldor, N. (1955) "Alternative theories of distribution", `Review of Economic Studies', 23, 83-100.
Kaldor, N. (1976) "Inflation and recession in the world economy", `Economic Journal, 86, 703-14.
Kalecki, M. (1954) "Costs and prices", reprinted in M. Kalecki (1971), `Selected essays on the dynamics of a capitalist economy', Cambridge: Cambridge University Press.
Lerner, A.P. (1933) "The concept of monopoly and the measurement of monopoly power", `Review of Economics
and Statistics', 1, 157-73.
Madhur, S. and P. Roy (1986) "Price setting in Indian industry", `Journal of Development Economics', 20, 205-24.
Mitra, A. (1978) `Terms of trade and class relations',
London: Frank Cass.
Nickell, S.J. and P. Kong (1987) "Wages, prices, employment and output", `Applied Economics Discussion Paper no. 29', Oxford: Institute of Economics and Statistics.
Raj, K.N (1976) "Growth and stagnation in Indian industrial development", `Economic and Political Weekly', 11, 223-36.
Rotemberg, R. and G. Saloner (1986) "A supergame theoretic model of price wars during booms", `American Economic Review', 76, 390-407.
Rowthorn, R. (1977) `Capitalism, conflict and inflation', London: Lawrence and Wishart.
Shapiro, C. (1989) "Theories of oligopoly behaviour", in `Handbook of Industrial Organisation' by R. Schmalensee and R. Willig, Amsterdam: North Holland.
Srinivasan, P.V. (1993) "Excess capacities in Indian industries: Supply or Demand determined", `Economic and Political Weekly', 27, 2437-41.
Srinivasan, T.N. (1993) "Demand deficiency and Indian industrial development", in P. Bardhan, M. Datta-Chaudhuri and T.N. Krishnan (ed.) `Development and Change: Essays in Honour of K.N. Raj, Delhi: Oxford University Press.
Srinivasan, T.N. and N.S.S. Narayana (1977) "Economic performance since the Third Plan and its implications for policy", `Economic and Political Weekly', 12, 225-39.
Stiglitz, J.E. (1984) "Price rigidities and market structure", `American Economic Review', Papers and Proceedings, 74, 350-5.
Taylor, J.B. (1979) "Staggered wage-setting in a macro model",
`American Economic Review', Papers and Proceedings,
69.
Taylor, L. (1983) `Structuralist macroeconomics', New York: Basic Books.
-------------------
Appendix
The data base of this study are the "Census of Manufacturing Industries" and the `Census sector' reports of the "Annual Survey of Industries". Thus data for the period 1950-58 come from the former source, while for the period 1959-82 they are drawn from the latter. The extent of the comparability of data from these two sources is discussed at length in Balakrishnan (1991). Limitations exist no doubt, but this procedure gives the longest time series of data on prices, costs, and activity in the manufacturing sector of the Indian economy. Restricting ourselves to the use of ASI data alone would yield a shorter series. This was treated as an important consideration when subjecting the data to standard econometric procedures.
Data definition and sources:
The charts are plots of the variables that are actually used in the estimation. Therefore the data in these exercises is uniform. The discussion of the definition and sources of the data that follows is grouped, broadly, according to whether they appear in the price equation or the output equation.
The price equation
Price: `P' is the wholesale price index, base 1970-71=100, for manufactured products. Sources are `Wholesale Price Statistics' by H.L. Chandok (1978, New Delhi: Economic and Scientific Research Foundation) for the period 1950-51 to 1977-78 and various issues of the `Report on Currency and Finance' (Bombay: Reserve Bank of India).
Apart from the dependent variable in the estimate of the price equation price variables appear in various forms in this study. The specific constructions are discussed below. However, the source of these data is uniform, being as mentioned above.
Labour cost: `L' = [E/O], where E stands for total emoluments, i.e., wages plus salaries plus benefits, and O is output. The amount of emoluments has been calculated from various reports of `CMI' and `ASI', respectively. Since the latter was not published for the year 1972, a simple average of the labour cost calculated for the years 1971 and 1973 was used.
Raw-materials cost: When a price equation is estimated for aggregate industry the representation of materials cost must exclude the value of manufactured inputs entering industry.
Thus `RM' is an index of the price of raw materials used by the manufacturing sector. This index was constructed as follows. The value of primary inputs into manufacturing output was used to construct shares of individual items in total materials costs. Inputs were classified into groups that could readily be represented by price indices. The shares of individual items in a group were aggregated to yield the weight assigned to the group in the composite index of raw material costs that was now constructed. The four groups and the respective weights are: `Food' - 2.4, `Non-food agriculture' - 77.5, `Minerals' - 12.2, `Fuel and power' - 7.9. The price indices chosen to represent these groups were drawn from the classification adopted in the official index number series, base 1970-71=100. These are `primary: food articles', `primary: non-food articles', `minerals', and `fuel, power, light and lubricants'. Sources: `Input-output Transactions Table (Commodity into Industry Absorption Matrix) for 1973-74', Central Statistical Organisation, for the derivation of the weights and the sources listed under `price' for the price series.
`Demand': Two indices have been used. `D' is the deviation of industrial output from its trend, and `Y' is the deviation of aggregate output (national income) from its trend. The index of industrial production has been used for the former. The source of this variable is discussed immediately below. The source for national income figures, used to construct the second index of demand, is `National Accounts Statistics', various issues.
`Money': The longest possible time series on the money stock in India was used. Source: A. Vasudevan (1980) "Money stock and its components in India, 1950-51 to 1979-80: A statistical account", `Indian Economic Journal', 28, 1-30.
The output equation
Output: `Xi' is represented by the index of industrial production. Splicing was undertaken to reduce it to a common base (1970-71=100). Source: `Economic Survey', various issues.
Public investment: `G ' is an index of capital formation by government. Source: `Capital formation and saving in India 1950-51 to 1979-80', Bombay: Reserve Bank of India.
Agricultural output: `Xa' is an index based on the triennium ending 1969-70=100. Source: `Agricultural Situation in India', February 1981.
Relative price: `RP' is the ratio of the index of the price of manufactures to the index of the price of agricultural goods.
See the discussion under `Price' for the source.