Productivity and efficiency results

The annual means of the Malmquist indices are presented in Table 8.4 below. These results show that over the 15-year period studied, there was a slight decline in the technical efficiency of the Australian natural gas distribution/retail industry. The figures in column 2 relate to the constant returns-to-scale case (pure technical and scale efficiency combined), and column 4 to the variable

Table 8.4 Annual means of Malmquist indices (1985/6-1998/9)

returns-to-scale case (pure technical efficiency). Under the constant returns-to- scale assumption, there was on average a 1.2 per cent annual decrease in technical efficiency in the performance of the natural gas distribution/retail industry. Assuming variable returns-to-scale in the industry, the annual decline was 0.8 per cent. Over the same period, scale efficiency fell by 0.4 per cent per year (column 5); however, for most of the years studied, the Australian gas industry experienced technological change (average technological change of 6.7 per cent per annum). High growth rates of technological change of over 10 per cent were estimated for 1985/6, 1994/5, 1997/8 and 1998/9. The overall result of the change in technological progress and efficiency was an increase in TFP of 5.4 per cent per annum for the industry (column 6). In terms of the success of the reform process, the mean of TFP change after 1991 was more than twice what it had been prior to this year (7.34 per cent versus 3.2 per cent).

As shown in Table 8.4 this growth in technological change indicates that during this period, the industry best-practice production frontier shifted significantly because of technological progress. Such improvement in technology led to increases in TFP in terms of greater output per given input. However, the different state-based gas sectors were not under the same pressure to improve their efficiencies, since some moved further away from the best-practice frontier.

Estimates for natural gas distribution/retail sectors in each state are presented in Table 8.5. Technical efficiency fell in New South Wales, South Australia and Queensland, and remained unchanged in Victoria and Western Australia. When taking into account variable returns-to-scale, Queensland appeared to be producing along the production frontier during the sample period. There was technological progress across all states, with New South Wales and Victoria recording the highest growth rates of 9.4 per cent and 12.1 per cent respectively. Correspondingly, TFP increased in all states.

A second-stage regression analysis can be performed on the Malmquist indices to detect the sources of efficiency differentials. In the context of this study, the indices were regressed on a number of exogenous variables, such as the specific

Table 8.5 Individual means of Malmquist indices (1985/6-1998/9)

Reform and productivity change 177

characteristics of the state-based natural gas sectors and their operating environments. The variables can be divided into two general groups. The first group of variables relates to specific characteristics of the sectors, such as the rate-of- return, the share of natural gas of the non-transport energy market and the share of the household energy market. Here it is assumed that the relationship between profitability and productivity is in the quadratic form with a one-year time lag; and that productivity is influenced by the growth rate of market share that the natural gas sectors held. Of particular interest in the Australia gas market context are the non-transport energy and household energy sectors respectively. The second group of variables is a set of dummy variables included to account for the differences among the sectors, such as the ownership type (government vs. private) and state locations. However, due to a lack of data, the observations on Queensland over the whole sample period and those of all states in the years 1985-7 and 1999 were dropped.⁴

The model is presented in the following equation (1), and the estimated results are further provided in Table 8.6 below:

where i = 1,. . .,4 and t = 1,. . .,11. The dependent variables in the three separate regressions are technological change (TE_it), technical change (TC_it) and total factor productivity (TFP_it) in each state (New South Wales, Victoria, Western Australia and South Australia), ranging from 1988 to 1998. Pr or(-1)_it, Nontron%_it, Household %_it are the explanatory variables representing the previous year’s rate-of-return for privately owned dominated sectors (0 used for publicly owned sectors), growth rate of market share of the non-transport energy sector and growth rate of market share of the household energy sector for natural gas sectors i at time period t. Ownership_it is the dummy variable for the ownership type of the state-based sectors (1 for predominantly privately owned sectors, and 0 otherwise). State1_it, State2_it, State3 _it are dummy variables for the states in which the sectors are located (State1 refers to New South Wales, State2 to Victoria, State3 to South Australia and State4 Western Australia). It has been assumed that ?_it may be subject to autocorrelation and heteroscedasticity.

The first four columns of Table 8.6 are the estimated coefficients, standard errors, t-statistics and the corresponding P-value for the regression of Malmquist technical efficiency on the vector of explanatory variables, after correcting for autocorrelation and heteroscedasticity. The test of the null hypothesis that all the slope coefficients are jointly zero is rejected at the 1 per cent level using Wald Chi-square statistics. As shown in the table, the gas sectors that are privately owned have a technical efficiency that seems to exhibit a U-shaped relation with their profitability in the previous year. Efficiency change first decreases, then rises with the rate-of-return in the previous year. Efficiency arrives at its

Table 8.6 Determinants of productivity variation

minimum when the annual rate-of-return is around 7.5 per cent. This indicates that well-performing sectors in terms of profitability have strong incentives to improve technical efficiency, while sectors that perform below average might be discouraged from making improvements or simply lack the capacity to do so. In addition, efficiency change over the sample period is higher for the gas sectors that are experiencing rapid expansion of market share for either the nontransport or household energy sectors. The ownership dummy is positively correlated with efficiency change at the 1 per cent significance level, with the coefficient of 0.2328. Privately owned sectors appear to be significantly more technically efficient than publicly owned. This is coincident with the fact that labor productivity increased sharply during the process of privatization in Western Australia, South Australia and Victoria.

The estimated coefficients, standard error, t-statistics and P-value for the regression of technological change on the vector of explanatory variables are presented in the following four columns of Table 8.6. A test of the null hypothesis of the joint insignificance of all the explanatory variables is rejected at the 1 per cent level, based on the conclusion that the vector of economic variables has a significant influence on the magnitude of technological changes. With relatively low explanatory power (adjusted R-squared is only 0.187), there are only two variables statistically significant at the 1 per cent level on technological changes (i.e. the ownership dummy and the Victoria state dummy). The ownership dummy is negatively related with technical change score, with coefficient of -0.1647. Private-owned sectors seem to have experienced less technological improvement than the public sectors. In addition, the state of Victoria appears to have experienced rapid technological progress, as it is positively related to technical change with a coefficient of 0.1599.

Other variables that may exert some influence on technological change are the rate-of-return in the private sectors, growth rate of market share of the household energy sector and the South Australia dummy variable, statistically significant at 7.3 per cent, 6.1 per cent and 8.7 per cent level respectively. The private sectors’ profitability is linearly rather than non-linearly related with the technological change scores. The coefficient of 0.0274 tells us that if the rate-of- return increases by 1 per cent, then the technological change score in the following year is expected to rise by 2.7 per cent. This might imply that profit-motivated private firms tend to make more investments in new technologies to achieve higher profits in the future. However, the negative ownership coefficient suggests that private sectors in general have experienced slower technological progress than public firms within the industry. The expansion of supplying gas to the household sector is also positively associated with technological change, with an estimated coefficient of 0.0154. The quicker expansion in terms of growth rate of market share that the sector has, the more rapid technological progress takes place. In addition to Victoria, South Australia appears to be another state where more rapid technological change took place during the sample period.

The final four columns of Table 8.6 are the estimated coefficient, standard errors, t-statistics and P-values where the dependent variable is the Malmquist

TFP index. The test of the null hypothesis that all the slope coefficients are jointly zero is rejected at the 1 per cent level. Like efficiency change, productivity change for the privately owned dominated sectors also has a U-shaped relation with its profitability in the previous year. It reaches its minimum when the annual rate-of-return is around 6.8 per cent. Both the estimated coefficients of the growth rate of market share and the growth rate of market share for household energy sectors are significantly positive at the 1 per cent level. This suggests that productivity improvements are generally higher for gas sectors experiencing rapid market expansion. However, in this study no clear relationship was found between ownership type and productivity. This might be due to the combined result of opposite effects of ownership on efficiency and technical change. It is also interesting to note that the sectors in New South Wales, Victoria and South Australia have often had relatively higher productivity change than those operating in Western Australia.