# Adjustment of Pricing: Evidence from Indian Manufacturing

Mita Bhattacharya, Michael Olive

#### This paper analyses the pricing behaviour in the Indian manufacturing sector, considering both domestic and external variables. Price adjustment models are developed based on industrial organisation literature and are examined with 28 manufacturing industries at the 3-digit level over the period from 1963 to 2001. Domestic structural factors are found to be important in determining the speed of price adjustment.

An earlier version of this research was presented at the 48th Annual Meeting of the Western Social Science Association, Phoenix, United States in April 2006.

Mita Bhattacharya (Mita.Bhattacharya@Buseco.monash.edu.au) is at the Department of Economics, Monash University, Australia and Michael Olive (molive@efs.mq.edu.au) is at the Macquarie University, Australia.

Economic & Political Weekly

may 23, 2009 vol xliv no 21##### 1 Introduction

ndustrialisation has been viewed as the engine of India’s growth since Independence. The ﬁrst phase of the liberalisation programme started in the mid-1980s, but gained m omentum after 1991. Economic reforms in the early 1990s r esulted in deregulation and the exposure of Indian industries to the domestic and international markets.Industrial development from the 1950s till the mid-1980s was regulated, with substantial control by government. Licensing and capacity controls in major industries led to concentrated market structures. Some sectors were reserved for small-scale ﬁrms in order to support unskilled or semi-skilled employment. Pricesetting behaviour by ﬁrms reﬂects competitiveness and may vary across industries and over time. With liberalisation, the role of international prices becomes important for domestic price-setting behaviour.

In a competitive market structure, the institutional framework ensures perfect price ﬂexibility. In a quasi-competitive industrial market, price adjustment by an individual ﬁrm may depend on its rivals. A seminal paper by Means (1935) has generated a large volume of literature on the effects of market structure on price adjustment. Domberger (1979, 1983) establishes a positive relationship between price adjustment and concentration (an indicator of market structure) in the case of UK manufacturing. On the other hand, Dixon (1983) reports a negative relationship for a newly industrialised country like Australia, which implies that the evidence in the literature is not very conclusive. This is due to the industrial structure and nature of the market in any particular economy.

The purpose of this paper is to explore new evidence on the relationship between price adjustment and market structure in a newly liberalised developing country such as India. Liberalisation of the economy has had a signiﬁcant inﬂuence on domestic market structure and prices. Reductions in tariffs, non-tariff barriers and licensing controls have given multinationals and foreign competitors access to the domestic market. Manufacturing in India is at a critical juncture. Most of the manufacturing ﬁrms are still well below world class practice. The introduction of foreign competition has simultaneously delivered reduced costs, improved quality, better performance, a wider range of products and better service. Given these changes, it is worthwhile analysing the structural determinants of the speed of price adjustment at the industry level. This research will reveal the consequences of these for price changes in Indian manufacturing.

The rest of the paper is set out as follows. Section 2 summarises the literature with major ﬁndings from related studies.

SPECIAL ARTICLE

Section 3 deals with the theoretical background explaining the determinants of price adjustment. Section 4 analyses the data and the methodology of the empirical investigation. Finally in Section 5 we summarise our major ﬁndings with some indicative remarks for policy purposes.

##### 2 Overview of Literature

Recent research on the speed of price adjustment links industrial organisation with macroeconomics. The consequences of imperfect competition on welfare have always been an important research question in industrial economics, while the effects of market imperfection have been investigated more recently. A popular method to add price dynamics into a pricing equation is by incorporating a term with the costs of price adjustment directly into a ﬁrm’s proﬁt function. As part of a project extending the microeconomic foundations of macroeconomics, Rotemberg (1982a,b) models these costs as a quadratic function. Martin (1993) starts from a proﬁt equation that incorporates a quadratic price adjustment cost function and takes the theoretical analysis a step further by deriving the speed of price adjustment as a function of market power. In the case of oligopoly, price adjustment is slower in concentrated industries.

Establishing a relationship between ﬁrm size and the speed of price adjustment is widely analysed in literature. Domberger (1983) suggests that large ﬁrms have large proﬁt cushions and, as a result, are less risk averse, leading to faster price adjustments. He concludes that “ﬁrms in concentrated industries with correspondingly higher price cost margins are more likely to behave as price leaders than those of more fragmented industries” (p 52).

A natural corollary to Domberger’s research is that ﬁrms have economies of scale over certain adjustment costs. This also happens to be a hidden assumption underlying quadratic price adjustment cost functions. In Section 3 of this paper, a model is derived that explicitly takes a positive relationship between economies of scale in regard to quadratic price adjustment costs and the speed of price adjustment. This assumes that ﬁrms with adjustment costs over a large range of output have less reason to slow their pace of price adjustment.

A second feature of the model is the negative relationship between market power and the speed of price adjustment, as discussed in Martin (1993). This proposition has some support in the empirical literature, as there are a number of studies that use industry concentration as a proxy for market power and ﬁnd a negative relationship with the speed of price adjustment (for example, Dixon 1983; Bedrossian and Moschos 1988; Weiss 1993 and Shaanan and Feinberg 1995 for Australian, Greek, Austrian and US manufacturing, respectively). Finally, averaging across the industry transforms the model into an error correction form, which places fewer restrictions on the short-run dynamics of the estimating equation when compared to a partial adjustment model.

A review of the empirical literature indicates that the length of the production period is also considered an important determinant of the speed of price adjustment. Domberger (1983) suggests that the valuation of inventories should be at historical cost, rather than opportunity cost, which sets up a disequilibrium wedge. This results in ﬁrms with short production periods placing a greater weight on the costs of disequilibrium, leading to

60 greater speeds of price adjustment. In support, Dixon (1983) ﬁnds a negative relationship between the speed of price adjustment and the production lag for Australian manufacturing industries.

In the case of India, the literature is still scarce; we describe here the existing studies. In a labour-abundant country like India, demand pressure has a signiﬁcant inﬂuence in determining pricing behaviour. In a study, Madhur and Roy (1986) develop models with variable mark-up rates for price adjustment and tested with annual data between 1961 and 1977 from four major sectors in India.1 The role of demand pressure and international prices are incorporated in a price setting model for both short- and long-run situations. They establish differences in lags in the adjustment of prices to costs amongst these sectors. Also, capacity utilisation (a proxy for demand pressure on prices) is found to have a signiﬁcant effect on mark-up, while international prices do not have much inﬂuence on the price-setting behaviour of ﬁrms.

Conversely, Chatterji (1989) analyses the behaviour of mark-up over the cycle for the aggregate and six individual industries over the period of 1949-77. For the aggregate industry price equation, demand was found to be an insigniﬁcant factor in determining prices.

Balakrishnan (1992) examines price cycle behaviour in relation to mark-up over the period 1952-80. Using an error-correction model, a price equation is developed for aggregate industry. Over a different range of activities, prices and costs are found to be cointegrated.

##### 3 Model Specification

In an imperfectly competitive industry suppose there are N ﬁrms, each ﬁrm producing a differentiated product. The short-run proﬁt function of the ith ﬁrm can be written as:

pit – pit–1

π(pit) = (pit – mcit) qit – αi ( )2 (q* ...(1)

it)S pit–1

where i and t represent ﬁrm and time subscripts, respectively, and pit, qit, q*it, mcit, αi and S indicate price, output, target output, constant marginal cost (excluding adjustment costs), a cost of adjustment parameter and an economies of scale parameter, respectively. The ﬁrst term on the right-hand side of (1) is revenue minus non-adjustment related costs, while the second term is the cost of price adjustment.2

When S is zero, the cost of price adjustment in (1) is the standard quadratic price adjustment cost function. This implies larger imposts on the ﬁrm for larger percentage price changes. Rotemberg (1982a) refers unfavourable customer reaction to higher prices as an example of this type of cost. Presumably, the ﬁrm imputes a value to the loss of current and future goodwill when prices are raised to levels above expectations or when prices are increased well in advance of competitor prices. In a similar but alternative scenario, ﬁrms uncertain about market conditions may be unsure ex ante that a given target price is optimal and so impute a cost to rapid price change (for a discussion, see Domberger 1983: 54-59).

Adjustment costs can also arise in input markets, with many authors pointing to turnover costs in relation to labour (see Kraft 1995; Kasa 1998; and Lindbeck and Snower 2001). Given the rationing role of prices, it seems reasonable to model these adjustment costs in the form of quadratic price adjustments under

may 23, 2009 vol xliv no 21

###### SPECIAL ARTICLE

c ertain conditions. For example, during a demand slowdown ﬁrms often hoard labour rather than face the costs associated with retrenching employees and then rehiring during the next upturn. A way of achieving this outcome and limiting losses in labour productivity is to maintain output levels through smaller price changes. In this paper, the quadratic price adjustment cost function is interpreted as representing an amalgam of implicit costs that can arise from adjustments in both product and input markets.

With the standard quadratic price adjustment cost function, the implicit cost to the ﬁrm of a given proportional price change remains the same regardless of ﬁrm size. Therefore, the absolute value of the cost of price adjustment would be the same for a multinational company as for a local artisan (given the same αi). This only makes sense if there are extreme economies of scale. In order to allow for varying scale effects, the price adjustment cost is also a function of the ﬁrm’s target output level. For a given price adjustment, it can be seen from (1) that the average cost of price adjustment declines with target output (economies of scale) when S is less than one; that it increases with target output (diseconomies of scale) when S is greater than one; and that it is constant when S is equal to one.

In the absence of adjustment costs, the ﬁrst-order condition for proﬁt maximisation is as follows:

q* + (p*it – mcit)(dq*it│dp*) = 0

itit ...(2) where * indicates the equilibrium values of price, output and the slope of the demand function. When adjustment costs are taken into consideration, q * it and p * it become the ﬁrm’s target output and target price, respectively (this assumption is standard in the literature). Given that the actual price and the target price differ, the ﬁrm’s output can be approximated using the following ﬁrstorder Taylor series:

≈q* + (dq*it│dp* )(pit – p* )

qitititit ...(3)

Substituting (3) into (1) explicitly expresses proﬁt as a function of price. After calculating the ﬁrst-order proﬁt maximising condition and incorporating (2) into the analysis, it can be shown that the ﬁrm chooses to change prices according to the following model:

∆pit = λit(p*it – pit–1)

...(4)

αiβit

λit = [1 – ( )]–1 ...(5)

1–S ηitp*itq*it

p*it dq* it ...(6)

ηit =q* dp*

itit

where ∆pit = pit – pit–1, λit is the speed of price adjustment, ηit is the elasticity of demand and βit = (p*it│pit–1)2. It is readily apparent that the range of λit is from zero to one and that (4) is just the partial adjustment model. Holding other things constant, it can be seen from

(5) that the ﬁrm’s speed of price adjustment increases/decreases with target output when the ﬁrm has economies/diseconomies of scale with respect to the costs of price adjustment; that the ﬁrm’s revenue is positively correlated with the speed of price adjustment when S is zero; and that as demand becomes more/less elastic the ﬁrm’s speed of price adjustment increases/decreases.

In order to give further direction to the empirical analysis in this paper, it is necessary to aggregate ﬁrm effects across the industry.

Economic & Political Weekly

may 23, 2009 vol xliv no 21Taking a weighted average of (4) across all ﬁrms in the industry and manipulating gives the following error correction model:

∆pdt = γdt∆p*dt – λdt(pdt–1 – δdtp*dt–1) ...(7)

where d is an industry subscript and ∆pdt = Σwi ∆pit, ∆p*

dt = Σwi ∆p*it, Σwi p*it λit Σwi pit–1 λit

pdt–1 = Σwi pit–1, p*dt–1 = Σwi p*it–1, γdt =, λdt =

# ()( )

Σwi p*it Σwi pit–1 and δdt = γdt│λdt. Following Bloch (1992), w1 represents the ith ﬁrm’s share of the value of industry shipments at a point in time. Therefore, the industry prices and target prices given in (7) are share-weighted averages. The error correction form of the model comes about because γdt and the industry speed of price adjustment (λdt) are differently weighted averages of each ﬁrm’s speed of price adjustment.3 If all ﬁrms in the industry have the same speeds of price adjustment, then the industry model reverts to the partial adjustment form.

In order to further inform the empirical analysis, the industry target price is derived when ﬁrms have log-linear and linear demand functions. The workings are shown in Appendix 1 (p 64). In the former case, the elasticity of demand is exogenous and the industry target price is a linear function of the weighted average of each ﬁrm’s marginal cost. In the latter case, the industry target price is a linear function of the weighted averages of each ﬁrm’s marginal cost and demand shift variables. Generally, pricing equations will be a function of cost and demand shift variables, except when the demand function is iso-elastic and moves p roportionally (Bloch 1992; Olive 2002).

##### 4 Data, Estimation and Empirical Finding

##### 4.1 Data

In this section we examine the determinants and stability of the industry speed of price adjustment for 28 Indian manufacturing industries at the three-digit International Standard Industry Category (ISIC) level during the period 1963 to 2001. Data used to construct series for industry price, industry average cost, manufacturing production, manufacturing price, competing foreign prices, average industry size and within industry competitiveness are obtained from IndiaStat, and from the United Nations Industrial Development Organisation (UNIDO) database. Additional import and export data used to construct measures of openness and import competition are taken from the International Economic Database (IEDB).4 Detailed descriptions of the data series and sources are in the Data Appendix (p 64).

##### 4.2 Empirical Test

Although cost plus pricing has been a common assumption in the industrial organisation literature since the survey work by Hall and Hitch (1939), there is plenty of evidence to suggest that industry price is a function of both demand and cost inﬂuences. Olive (2002, 2004) estimates a pricing equation for manufacturing industries in 11 countries and ﬁnds that while average variable cost is the dominant determinant of industry price, for all countries demand variables are also signiﬁcant. In the case of India, average variable cost, competing foreign price and aggregate manufacturing price are found to inﬂuence industry price signiﬁcantly.

SPECIAL ARTICLE

This supports the ﬁnding of variable mark-up rates for Indian industry by Madhur and Roy (1986).

Consistent with these ﬁndings, industry target price is modelled as a linear function of industry average cost (acdt), manufacturing production (imandt), manufacturing price (pmandt), Japanese industry price (jpdt) and US industry price (uspdt). It should be noted that manufacturing production and manufacturing price vary across industries as they exclude the dth industry’s own output and price. Also, the industry prices for Japan and the US are included as proxies for competing foreign price, as no such series is available for India.

Given the error correction model developed above, the basic empirical pricing equation for each industry is of the form:

Δpdt = θd1 + θd2Δacdt + θd3Δimandt + θd4Δpmandt + θd5 Δjpdt + θd6 Δuspdt

– λdECMdt–1 + εdt ...(8)

– – imandt–1 – φd4pmandt–1 –

ECMdt–1 = pdt–1φd2acdt–1φd3φd5jpdt–1...(9)

– φd6uspdt–1 where Δindicates ﬁrst difference, θd2 to θd6 are short-run parameters, θd1 and φd2 to φd6 are long-run parameters, ECMdt-1 is the error correction mechanism, and εdt is an error term. Also, the expected signs

are shown in (8) and (9). Initially, the industry speed of price adjustment (λd) is taken to vary across industries but not across time.

##### 4.3 Panel Estimations and Related Testings

A consequence of non-stationary errors can be a spurious OLS regression that over-rejects null hypotheses. In order to test whether the error term is indeed stationary, the time series properties of the data are investigated, with the results presented in Table 1. The Im, Pesaran and Shin (IPS 2003) test for unit roots in panel data indicates that industry price, industry average cost, manufacturing production, manufacturing price, Japanese i ndustry price and US industry price each have a unit root in levels but are ﬁrst-difference stationary. A maximum of eight lags and a time trend are allowed for in all these time series and the results hold at the 1% level of signiﬁcance. If these variables in levels are

Table 1: Tests for Non-Stationarity of Series in Natural Logarithm Form

Variable | Level | First Difference | Test Type |
---|---|---|---|

pdt | 1.54 | -11.72** | IPS |

acdt | 2.06 | -12.92** | IPS |

imandt | 8.16 | -19.12** | IPS |

pmandt | 1.89 | -12.98** | IPS |

jpdt | 3.43 | -12.76** | IPS |

uspdt | 4.95 | -7.70** | IPS |

cointegration | -3.77** | Group ADF |

Panel data tests are carried out using Pedroni programme for RATS. IPS indicates Im, Pesaran and Shin (2003) test for unit roots in panel data. Group ADF indicates Pedroni (1999) test for cointegration in panel data. For each test the null hypothesis is non-stationarity. The panel data test statistics are z distributed under the null and all unit root tests have a maximum eight lags and a time trend, while the test for cointegration has no time trend. ** indicates significant at the 1% level for a one-tailed test.

* indicates significant at the 5% level for a one-tailed test.

cointegrated, then the error correction mechanism and the error term are both stationary. Using Pedroni’s (1999) Group ADF test for panel data, the null of no cointegration is rejected at the 1% level of signiﬁcance. Therefore, the results presented in Table 1 give us conﬁdence in the inferences we make.

The industry speed of price adjustment in (8) is estimated using a two stage procedure. The ﬁrst stage is to estimate the parameters of the error correction mechanism for each industry. When the appropriate short-run dynamics are excluded, Patterson (2000) shows that long-run parameter estimates may be biased for ﬁnite series and Kremers et al (1992) show that hypothesis tests of the speed of adjustment are likely to have low power. Therefore, (9) is estimated for each industry using non-linear least squares and allowing for an additional lag in the short-run difference variables. In the second stage, industry speeds of price adjustment are obtained by estimating (8) as a system of equations (with ECMdt-1 calculated from the ﬁrst stage) using the method of dummy variable least squares (DVLS).

With reference to the economic model, industry speed of price adjustment is modelled as a linear function of average ﬁrm size and variables that are likely to affect the industry elasticity of demand. In the ﬁrst instance, we take the traditional approach and assume that structural variables that inﬂuence the speed of price adjustment change across industry, but not (rapidly) across time (for example, see Domberger 1983; Dixon 1983; Bedrossian and Moschos 1988; Kardasz and Stollery 1988; Weiss, 1993; Shannan and Feinberg 1995). By regressing the industry speed of price adjustment estimates on these structural variables we can assess their inﬂuence.

Given economies of scale with regard to price adjustment costs, it is expected that average ﬁrm output will be positively correlated with the industry speed of price adjustment. Here, industry output divided by the number of establishments for 1981 is used to represent average ﬁrm size (SIZd).

With regard to heterogeneous goods, Sawyer (1982) suggests that industry concentration may act on ﬁrm price conjectures, and resulting in slower price adjustment. In the case of Indian manufacturing, major sectors were controlled by government (examples include steel, aviation, petrochemicals, automobiles) with licensing requirements and capacity control. Before the 1990s, manufacturing prices were quite rigid. One would expect that an increase in industry concentration would decrease the industry speed of price adjustment. We use the inverse of the number of establishments (Nd) for 1981 as a proxy, i e, we expect a positive relationship between the number of establishments and the industry speed of price adjustment.

In a formal model, Bloch (1994) shows that higher import shares reduce the elasticity of price conjectures, thus making demand more elastic. In light of our model, we expect greater import competition to increase the speed of price adjustment. The measure of import competition (MSHd) employed is the value of industry imports divided by the value of industry output averaged over the 1963-2001 period.

Table 2 (p 63) shows the results from the regression of the industry speeds of price adjustment on SIZd, Nd and MSHd and on their square roots SIZRd, NRd and MSHRd using OLS and WLS. In the latter case, the standard errors from the estimated speeds of price adjustment are used as weights. It can be seen that there is some support for our economic model as SIZd and SIZRd are signiﬁcantly positive at the 5% level for OLS and signiﬁcantly positive at the 10% level for WLS, while Nd and NRd are signiﬁcantly positive at the 5% level and 1% level, respectively, for WLS. Notably, the results also show that import competition is not signiﬁcant in any of the regressions.

may 23, 2009 vol xliv no 21

SPECIAL ARTICLE

Thus far, our investigation suggests that particular cross sectional variables are important determinants of industry speed of price adjustment. However, a different approach to that taken for Table 2 is necessary in order to examine the impact of time

Table 2: Regression of Industry Speeds of Price Adjustment on Various Structural Variables

Estimation | Structural Variables | ||||

Method | SIZd Nd | MSHd | SIZRd | NRd | MSHRd |

OLS | 3.15* -1.10 | -0.13 | |||

(2.60) (-0.68) | (-0.42) |

OLS | 1.78* | -0.39 | -0.13 | |||

(2.48) | (-0.17) | (-0.42) | ||||

WLS | 4.91# | 4.96* | -0.04 | |||

(1.79) | (2.35) | (-0.31) |

WLS 2.82# 6.80** 0.06

(2.03) (3.21) (0.47)

t-statistics are in parentheses and there are 28 observations. Parameter values for SIZ, SIZRd,

d

N and NR should be multiplied by 10-10, 10-5, 10-5, and 10-3, respectively, when firm size is

dd

expressed in rupees and N is the absolute number of firms in an industry.

d

OLS indicates estimation by ordinary least squares. WLS indicates weighted least squares, where the standard errors from the estimation of the industry speeds of price adjustment are used as weights. ** indicates significant at the 1% level for a two-tailed t test. *indicates significant at the 5% level for a two-tailed t test. # indicates significant at the 10% level for a two-tailed t test.

varying inﬂuences on industry speed of price adjustment. During the considered period, signiﬁcant changes have taken place including the initiation of economic reforms in the early 1990s. The new pricing equation is given as:

= θd1 + θd2Δacdt + θd3Δimandt + θd4Δpmandt + θd5Δjpdt +

Δpdtθd6Δuspdt – (λ1 + λ2SIZd – λ3Nd + λ4Vt )ECMdt–1 + εdt ...(10)

where λ1 to λ4 are parameters constrained to be the same across industries and V represents a time varying determinant of indus

t

try speed of price adjustment. Estimation of (10) now involves directly entering the determinants of the speed of price adjustment into the second stage and constraining their coefﬁcients to be the same across industries.

However, estimating (10) introduces a bias into the estimates of λ1 to λ4 when prices are in index form. If a price index for an industry can be characterised as an unknown number multiplied by the true price, then constraining the speed of price adjustment across industries is the same as multiplying the dependent variable and the error correction mechanism in (10) by an arbitrary unknown variable. This problem can be overcome to some extent by transforming pdt, acdt, imandt, pmandt, jpdt and uspdt into natural logarithms. Under the null hypothesis that λ4 is zero, estimation then leads to unbiased estimates of λ1 to λ3 and the unknown numbers simply fall out into the constant ﬁxed effects for each industry. By similar reasoning, any non-zero estimate of λ4 is likely to be biased when the above variables are transformed into natural logarithms.

We proceed by initially estimating the parameters for the error correction mechanism and for (10) assuming that λ4 is zero. Then we move to test a range of variables that have changed relatively rapidly over time in order to see whether our null position is rejected. The ﬁrst of these variables is a reform dummy variable (RD) that is zero from 1963 to 1990 and is one from 1991 to 2001. This is meant to capture the impact of the wide-ranging liberalisation of the Indian economy on the industry speed of price adjustment. A more speciﬁc variable that has been inﬂuenced by these reforms is the degree of openness in each industry (OPdt). This is measured as imports plus exports divided by imports plus domestic output. If

Economic & Political Weekly

may 23, 2009 vol xliv no 21either of these variables increases competitiveness in the economy, then we would expect them to increase the industry speed of price adjustment. A third variable is import share (MSHdt), which we now allow to vary across time and industry. Finally, the growth in manufacturing output (MOdt) is employed to capture the inﬂuence of the business cycle on industry speed of price adjustment.

Table 3 shows that SIZd and Nd have the expected positive sign and Nd is signiﬁcant at the 1% level, but that SIZd is no longer signiﬁcant. By multiplying the coefﬁcient estimate by the actual

Table 3: DVLS estimation Results for the Industry Speed of Price Adjustment When It Is a Function of Cross-Sectional and Time Variables

Results When Time Series (That Are Not Part of λd) Are in Natural Logs | |||||
---|---|---|---|---|---|

Variable | (1) | (2) | (3) | (4) | (5) |

Constant | 0.45** | 0.46** | 0.45** | 0.45** | 0.50** |

(9.46) | (9.58) | (9.52) | (9.49) | (6.70) |

SIZd | 0.34 (0.53) | 0.28 (0.44) | 0.29 (0.44) | 0.31 (0.49) | 0.28 (0.42) |

Nd | 3.56** | 3.45** | 3.54** | 3.55** | 3.51** |

(3.31) | (3.23) | (3.31) | (3.30) | (3.22) | |

RD | 0.01 | ||||

(1.01) | |||||

OPdt | 0.02# | ||||

(1.71) |

MSHdt | 0.01 | |
---|---|---|

(1.50) | ||

MOdt | -0.82 | |

(-0.86) |

t-statistics computed from heteroscedastic-consistent standard errors are in parentheses. Parameter values for SIZ, and N should be multiplied by 10-10 and 10-5, respectively, when firm

dd

size is expressed in rupees and N is the absolute number of firms in an industry.

d

** indicates significant at the 1% level for a two-tailed t test. # indicates significant at the 10% level for a two-tailed t test.

values we can get an idea of the contribution of SIZd and Nd to the speed of price adjustment. For formulation (1), the maximum contribution by ﬁrm size is 0.10 in petroleum reﬁneries (353) and the maximum contribution by the number of ﬁrms is 0.65 in food products (311). Formulations (2), (3) and (4) show positive estimates for RD, OPdt and MSHdt, suggesting that the economic reforms have increased the speed at which ﬁrms adjust to their long-run equilibrium. However, only openness is signiﬁcant at the 10% level and its contribution to the industry speed of price adjustment appears to be modest (as discussed above, the estimate for OPdt may be biased). It can also be seen from Table 3 that manufacturing output growth is not signiﬁcant, which suggests that industry speed of price adjustment is not affected greatly by the business cycle.5

##### 5 Implications and Conclusions

The purpose of this paper was to analyse the speed of price a djustment in the case of Indian manufacturing. An error correction model is considered in explaining adjustment. Industry average cost, manufacturing production and price, and international price are used as independent variables. Both cross section and time-varying effects are considered in explaining the model. Among cross sectional variables, average ﬁrm size is found to have a positive signiﬁcant effect on speeds of price adjustment for all speciﬁcations. There is some support for the view that concentration has a positive and signiﬁcant effect on the speed of price adjustment. The inﬂuences of time variables in our model have second order effects. This implies that the reform programme and other time variables inﬂuence prices through their effects on the target price. Concentration is still a signiﬁcant determinant of the speed of price adjustment. This is due to the protected

SPECIAL ARTICLE

Indian industrial structure which has led to restrictive environ-liberalised markets are increasing the speed of price adjustment. ment to compete in the world market. We could not establish that In summary, our ﬁndings suggest that the domestic market output growth in manufacturing is affected by the business cycle. structure is important in determining price adjustment both This implies that the industry speed of price adjustment is not across industries and over time. Also, recent liberalisation has inﬂuenced greatly by the business cycle. Also globalisation and had some effects on price adjustment.

Notes

1 The four sectors are: capital goods, consumer goods, intermediate goods and basic goods. 2 Zero ﬁxed costs are assumed for simplicity. This does not affect the analysis.

3 This method for obtaining an error correction model could be contrasted with those outlined by Nickell (1985).

4 IEDB, Research School of Paciﬁc and Asian Studies, Australian National University.

5 The authors obtained similar results to those in Table3 when SIZRd, NRd and MSHRd are employed in the estimating equations.

References

Balakrishnan, P (1992): “Industrial Price Behaviour in India: An Error-correction Model”, Journal of Development Economics, 37, 309-26.

Bedrossian, A and D Moschos (1988): “Industrial Structure, Concentration and the Speed of Price Adjustment”, Journal of Industrial Economics, 26, 459-75.

Bloch, H (1992): “Pricing in Australian Manufacturing”, Economic Record, 68, 365-76.

– (1994): “The Impact of Foreign and Domestic Competition on Pricing in Australian Manufacturing” in M Johnson, P Kriesler, A Owen (ed.), Issues in Australian Economics (North Sydney: Allen and Unwin)

Chatterji, R (1989): The Behaviour of Industrial Prices in India (Oxford University Press).

Dixon, R (1983): “Industry Structure and the Speed of Price Adjustments”, Journal of Industrial Economics, 32, 25-27.

Domberger, S (1979): “Price Adjustment and Industrial Structure”, The Economic Journal, 89, 96-108.

– (1983): Industrial Structure, Pricing and Inﬂation (Oxford: Robinson and Company). Hall, R L and C J Hitch (1939): “Price Theory and Business Behaviour”, Oxford Economic Papers, 2, 12-45.

Im, K S, Pesaran, and M H, Shin, Y (2003): “Testing of Unit Roots in Heterogeneous Panels”, Journal of Econometrics, 115, 53-74.

Kasa, K (1998): “Identifying the Source of Dynamics in Disaggregated Import Data”, Journal of Applied Econometrics, 13, 305-20.

Kardasz, Stanley W and Kenneth Stollery (1988): “Price Formation in Canadian Manufacturing Industries”, Applied Economics, 20, 473-83.

Kraft, K (1995): “Determinants of Price Adjustment”, Applied Economics, 27, 501-07.

Kremers, J J M, N R Ericsson and J J Dolado (1992): “The Power of Cointegration Tests”, Oxford Bulletin of Economics and Statistics, 54, 325-48.

Lindbeck, A, D J Snower (2001): “Insiders versus Outsiders”, Journal of Economic Perspectives, 15, 165-88.

Madhur, S and P Roy (1986): “Price Setting in Indian Industry”, Journal of Development Economics, 20, 205-24.

Means, G C (1935): “Industrial Prices and Their Relative Inﬂexibility” (Washington: US Senate Document, 13, 74th Congress, 1st Session).

Martin, C (1993): “Price Adjustment and Market Structure”, Economics Letters, 41, 139-143.

Nickell, S (1985): “Error Correction, Partial Adjustment and All That: An Expository Note”, Oxford Bulletin of Economics and Statistics, 47, 119-29.

Olive, M (2002): “Price and Markup Behaviour in Manufacturing: A Cross-country Study” (Chel en-

Chel

enham, UK: Edward Elgar).– (2004): “Speed of Price Adjustment and Price Conjectures”, Macquarie Economics Research Papers 2/2004.

Patterson (2000): “An Introduction to Applied Econometrics: A Time Series Approach” (Houndmills, UK: Macmillan).

Pedroni, P (1999): “Critical Values for Cointegration

Tests in Heterogeneous Panels with Multiple

Regressors”, Oxford Bulletin of Economics and

Statistics, 61, 653-70. Rotemberg, J J (1982a): “Sticky Prices in the United

States”, Journal of Political Economy, 90, 1187-1211.

– (1982b): “Monopolistic Price Adjustment and Aggregate Output”, Review of Economic Studies, 49, 517-31.

Sawyer, M C (1982): “On the Speciﬁcation of Struc

ture Performance Relationships”, European Eco

nomic Review, 17, 295-306.

Shaanan, J and R M Feinberg (1995): “Dynamic Com

petition and Price Adjustments”, Southern Eco

nomic Journal, 62, 460-66.

Weiss, C (1993): “Price Inertia and Market Structure: Empirical Evidence from Austrian Manufacturing”, Applied Economics, 25, 1175-86.

Appendix 1

Derivation of the industry target price when ﬁrms have log-linear demand functions and linear demand functions. Suppose, the log-linear demand function of the ith ﬁrm is:

log qit = Ait + ηi log pit ...(A1)

where Ait is a function of demand shift variables and ηi is the exogenous elasticity of demand. Rearranging the ﬁrst-order condition as presented in (2), the ﬁrm’s target price can be written as:

p* = (1 + 1/ ηi)–1mcit ...(A2)

itIt can be seen from (A2) that the target price is a linear function of marginal cost at the ﬁrm level. Taking a weighted average across ﬁrms, the industry target price can be written as:

p*dt = edmcdt ...(A3)

where p* = p* = p* and

dtΣwi it, mcdtΣwi it Σwimcit(1 + 1/ ηi)–1

ed = . Given that the elasticity

Σwimcit

of demand and marginal cost are uncorrelated in this case, E(ed) = Σ (1 + 1/ ηi)–1/N where E is the expectations operator and N is the number of ﬁrms in the industry.

Now suppose the ith ﬁrm has the following linear demand function: qit = Ait – bipit ...(A4)

where bi is a parameter that incorporates the ﬁrm’s price conjectural variations with regard to other ﬁrms in the industry (for the impact of price conjectures entering in this manner on the partial adjustment model, see Olive 2004). The ﬁrm’s target price is obtained by substituting (A4) into the ﬁrst-order condition and rearranging to give:

Ait mcit p*it = 2bi + ...(A5)

2

Equation (A5) represents the target price as a linear function of marginal cost and the demand shift variables. Taking a weighted average across ﬁrms, the industry target price can be written:

b–1 Adt mcdt p*dt = d+

2 2

may 23, 2009

Σwibi–1Ait–1–1

where bd = Σwibi and = .

Adt–1

Σwibi

Therefore, the industry target price is a linear function of the average inﬂuence of demand shift variables and marginal cost on the ﬁrm’s target price.

###### Data Appendix

UNIDO data are used to construct the variables pdt, acdt, imandt, pmandt, jpdt, uspdt, SIZd, Nd and MOdt, while additional import and export data from IEDB are used to construct OPdt and MSHdt. The data in this study are at the threedigit ISIC level. pdt – The Indian industry price is constructed by dividing gross output by the index of production. A similar process is employed to construct price indices for Japan and the US. However, jpdt and uspdt are converted into rupees by multiplying by the average annual rupee/yen and rupee/dollar exchange rates, respectively acdt – Industry average variable cost is constructed by subtracting value added from gross output plus wages and salaries, and dividing this by the index of production. imandt – The manufacturing production index for a particular industry is an output chain weighted average of the indices of production for the other 27 industries. The output weights are changed for the years 1963, 1970, 1980 and 1990. pmandt – The manufacturing price index for a particular industry is an output chain weighted average of the price indices for the other 27 industries. The output weights are changed for the years 1963, 1970, 1980 and 1990. Nd – The number of establishments in each industry in 1981. SIZd – The average establishment size is

obtained by dividing industry output by the number of establishments in 1981. MSHdt – Import competition is obtained by dividing imports by output in the same industry category. Imports are only available on an annual basis from 1970 to 1996. Therefore, the values going forward from 1997 are an average of the 1991 to 1996 import competition values and the values going back from 1969 are an average of the 1970 to 1974 import competition values. OPdt – Openness is obtained by dividing imports plus exports by imports plus output in the same industry category. Imports and exports are only available on an annual basis from 1970 to 1996. Therefore, the values going forward from 1997 are an average of the 1991 to 1996 openness values and the values going back from 1969 are an average of the 1970 to 1974 openness values. MOdt – Growth in manufacturing output is the proportional rate of change in the manufacturing production index for a particular industry.

vol xliv no 21