STOCHASTIC FRONTIER ANALYSIS OF INDONESIAN FIRM EFFICIENCY: A NOTE

This research attempts to model performance measurement for the firms listed on Indonesia Stock Exchange (IDX) using the stochastic frontier approach. There are 121 firms analyzed over the period of 2000-05 with 726 pooled observations. We also test whether firm’s age, size, market share, manufacturing classifications and time period have effects on the technical inefficiency of the manufacturing sector. Our findings reveal that the average technical efficiency of the tested firms is 0.7149, which is below the efficiency frontier: factors that affect inefficiency are found and explained. Our research has offered notable original contributions to performance measurement and provides insights on managerial decision making on operational performance of listed firms in an increasingly competitive Indonesian economy.


Introduction
Prior research on the Indonesian economy used stochastic frontier analysis (SFA) for evaluating a firm's performance but on other than manufacturing sectors: agriculture (see Daryanto, Battese, and Fleming (2002)) on technical efficiencies of rice farmers in West Java; Public and Private Sectors (Viverita and Ariff (2004); Viverita and Ariff (2006)); commercial banks (Abidin and Cabanda (2007); electronics manufacturing plants (Palangkaraya and Yong (2006); manufacturing sector's labor growth (Jacob and Los (2006)); and consumer industry (Probowo and Cabanda (2010).However, those few studies on Indonesian manufacturing firms are not (2001) applied a stochastic frontier approach to Korean manufacturing industries and showed that technical efficiency had a significant positive effect on its productivity growth.In another study, Söederbom and Teal (2001) examined three dimensions of the performance of firms in Ghana's manufacturing sector.The findings of these previous studies will be later compared to the new empirical findings derived in this research.
Our research attempts to model performance measurement for the firms listed on the IDX.This research has three specific objectives: (1) Determine the stochastic frontier measures on labor, inventory, fixed assets, and capital on total sales; (2) Test whether firm's age, size, market share, manufacturing classifications, and time period have effects to the technical inefficiency of the sector; and (3) Test whether there is a significant difference among technical efficiency (TE) scores of classifications.New findings will offer significant and new empirical contributions to the performance management field.
The rest of the paper is structured as follows.Section 1 discusses the state of the Indonesia's manufacturing sector.The economic and regulatory environment is described in section 2. Data, variables and the model are presented in Section 3 as part of the methodology.Section 4 presents new findings and our discussion while the conclusion and managerial implications research is in the last section.

Overview of Sector Studied
From the late 1970s, Indonesia experienced a rapid economic growth which was sustained over the next three decades.The economy was transformed from highly dependent on agriculture in 1960s into one in which this sector's contribution was more than a quarter of the gross domestic product (GDP) in the mid-1990s.From 1973 to 1980, the value of Indonesian export was dominated by oil/gas and timber (60 per cent).Later on, as more and more processing plants developed domestically, the share of semi-processed goods in total exports rose steadily and the in the mid 1990s became one of the most important foreign exchange earners.The 1997 financial crisis turned the economic miracle into shambles.By January 1998 the currency had depreciated by 80 per cent, while the economy contracted sharply to 51 per cent of GDP at its trend growth.
With the loss of valuable times, as the confidence of public and investor continued to evaporate, the crisis that was relatively mild in October 1998 continued to deepen when financial crisis led to political one.The currency continued to slide and the crisis had serious consequences; output contracted by 51 per cent of GDP, and US $ 238.60 billion estimated cost of the crisis (Widianto et al., 2000).
The severe economic contraction in 1998 was slightly reversed in 1999, when the economy grew again, though at a miniscule rate of 0.8 per cent.Rupiah was stabilized around Rp 9,000 per US dollar since November 2002a far cry from its 3,000 rupiah to dollar during pre-crisis times.The appreciation of Rupiah from around 15,000 rupiah along with the availability of food supply has held inflation in check.Measured by consumer price index (CPI), inflation reached its peak in 1998 at 82 per cent per annum.The inflation rate in 2001 was 11.2 per cent, 10.0 per cent (2002), 5.1 per cent (2003), 6.4 per cent (2004), 17.1 per cent (2005), and less than 10 per cent (2011).The inflation rate was higher in 2005 due to government decision to increase gas and oil prices by 100 per cent in 2005 (BPS Statistic Indonesia, 2006).
From 2000 through 2003, economic growth was mainly driven by private and public consumption, while fixed investment, just like in the preceding years after the crisis, remained sluggish.As a result of sluggish investment growth, the investment to GDP ratio in 2003 dropped to 17.8 per cent in 2003, the lowest level since the early 1970s.During the late Soeharto era, the investment to GDP ratio was around 30 per cent.However, in 2004 for the first time after the Asian crisis, GDP growth just exceeded 5 per cent.This time growth was not only driven by consumption, but also by investment, the growth of which for the first time after the crisis grew at double digit at 15.7 per cent.Export growth at 8.5 per cent was also higher than in 2002 and 2003.During the first and second quarters of 2005 fixed investment continued its double-digit growth (Wie, 2006).
The manufacturing sector accounts for an increasing share of GDP.The manufacturing sector accounted for an estimated 27.6 per cent of GDP in 2001GDP in , 27.8 percent (2002GDP in ), 28.0 per cent (2003GDP in ), 28.36 per cent (2004GDP in ), and 28.1 per cent (2005) ) of GDP: it is close to a third in 2011.The growth rates were 3.8 per cent (2001), 5.3 per cent (2002), 5.3 per cent (2003), 6.4 per cent (2004), and 4.6 per cent (2005).The sector contributes the highest contribution to Indonesian GDP growth from the year 2001 to 2005 (BPS-Statistic Indonesia, 2006).With this, the financial sector has responded well with its own rapid growth and rehabilitation to a healthy state.
In the late 1980s and the 1990s, Indonesia implemented policies designed to move toward a freer, more market-oriented financial system.Indonesia deregulated its financial sector in 1988-1989.There were 56 listed companies before the deregulation of the financial sector in 1988-1989. One year later (1990)), there were 123.Subsequently, there were 349 listed firms as of December 2005.In the manufacturing sector, there are 127 firms listed on Jakarta Stock Exchange (JSX).The JSX changed its name to Indonesia Stock Exchange (IDX) in December 2007.These firms listed are categorized into three classifications: basic industry (48 companies), consumer goods industry (38 companies), and miscellaneous industry (41 companies).

Data Sample
This research covers 121 out of the total 127 manufacturing firms listed on IDX from 2000 to 2005: due to data unavailability for recent periods, 2005 financial reports are the latest available.
A pooled data of 726 represent the panel data for the current analysis.Data were gathered from audited annual financial reports of manufacturing firms from Securities and Exchange Commission (BAPEPAM) and IDX.This research include all the three listed manufacturing classifications: basic industry (47 companies), consumer goods industry (36 companies), and miscellaneous industry (38 companies).All financial data were adjusted for inflation, using the Consumer Price Index (CPI) with a base year of 1993 prices.

Stochastic Frontier Analysis Model
We attempt to propose a model for technical inefficiency effects in a stochastic frontier production function for panel data of the listed manufacturing firms to estimate the trans-log stochastic production function over the time period.Provided the inefficiency effects are stochastic, the model permits the estimation of both technical change in the stochastic frontier and time-varying technical inefficiencies.Battese and Coelli (1995) provided the stochastic frontier production function for panel data: where Y it denotes the production at the t-th observation (t = 1,2, …,T) for the i-th firm (i = 1,2, …,N); it x is a it z is a ) 1 ( xm vector of explanatory variables associated with technical inefficiency of production of firms over time; and  is a ) 1 (mx vector of unknown coefficients (Battese and Coelli (1995).
To characterize the stochastic frontier production of the listed manufacturing sector firms, this research applies a trans-log stochastic production function.Applying the Battese and Coelli's (1995) where: Y it represent total sales of the manufacturing firm i-th at the t-th year of observation; it I represent inventory of the manufacturing firm i-th at the t-th year of observation; it F represent fixed assets of the manufacturing firm i-th at the t-th year of observation; it K represent capital of the manufacturing firm i-th at the t-th year of observation; it L represent labor of the manufacturing firm i-th at the t-th year of observation; Furthermore, Battese and Coelli (1995), specified the technical inefficiency effect, it U , in the stochastic frontier model as shown in Equation (3): where Age it represents the number of operation years of the manufacturing firm i-th at the t-th year of observation; Size it represents the total assets of the manufacturing firm i-th at the t-th year of observation; it e Marketshar represents sales of the manufacturing firm i-th at the t-th year of observation divided by total sales of the manufacturing sector; Class it represents the classification of the manufacturing firm i-th at the t-th year of observation; Time period it represents the time period of the manufacturing firm i-th at the t-th year of observation (2000 -2005); and it W is defined by the truncation of the normal distribution with zero mean and variance.
The stochastic frontier production function may investigate a firm's technical efficiency and may also identify factors for the technical inefficiency effects of the manufacturing sector firms.The computer software known as Frontier 4.1 by Tim Coelli was used to derive all empirical findings in this research.

Empirical findings
The value of the generalized likelihood-ratio (LR) statistics for the parameters in the stochastic production function for sales is shown in Table 2.The null hypothesis that that the Cobb-Douglas functional form is a correct functional form to represent the data in Indonesia's listed sector is significantly rejected.Therefore, the trans-log model is chosen based on the LR value of 155.59.This is greater than the critical value of 18.30 based on a Chi-square distribution table, tested at 5 per cent probability level.The null hypothesis that there is no technical inefficiency effect in the model is also significantly rejected, based on the LR value of 546.26, implying that inefficiency effect is present in the model.(Kodde and Palm, 1986).

Panel I Findings
To determine the stochastic effects of labor, inventory, fixed assets, and capital on total sales, results are shown in Table 3.The estimated coefficients of four inputs for the sector are reported in Panel I.There are five coefficients out of 14 that are significantly different from zero at the 5 per cent probability level.One direct effect, three squared terms and one cross product have coefficients significantly different from zero.These findings support the rejection of the Cobb-Douglas model: this is not an adequate representation of the sector.Inventory, among the four inputs, remains the single most significant predictor of sales output (efficiency), with an estimated elasticity of 0.7182.Overall, constant ( 0  ) is statistically significant (4.0894).This finding suggests that the joint effects of four predictors of technical efficiency in this sector are positive and significant, in general, while individual effects of one or more variables are not statistically significant.Labor shows a negative effect (-0.2799) but is statistically insignificant.This finding is consistent with the findings of Wei Koh et al. (2004) and Gholami, Moshiri, and Yong (2004); they found out that technical efficiency decreases as more labor inputs are used.Inventory coefficient has the estimated coefficient for fixed assets (0.1511), which is positive, but the effect is insignificant.
Lastly, capital (-0.0241) is found to have a negative but insignificant effect on efficiency, suggesting that efficiency declines when more capital is injected.This result supports the finding of Lundvall and Battese (2000) on Kenyan industry.This results are indicative of the sector's lack luster productivity.

Panel II findings
To further test whether firm's age, size, market share, classifications, and time period have effects on technical inefficiency of the sector, and the findings are shown in Table 3, Panel II.
Overall, the joint effect of five z-variables on the technical inefficiency is significant, where the constant is -24.6063.The estimated coefficient associated with age (0.1938) is positive and statistically significant, suggesting that older firms are technically inefficient than younger firms perhaps due to the latter adopting newer technology.Size is also found to have a positive significant effect on technical inefficiency, which is a normal results.This finding is consistent with the results of Biggs et al. (1996) that larger firms are technically inefficient than smaller firms.
Meanwhile, market share is found to have a negative effect on technical inefficiency and is statistically significant.This finding supports Tybout (2000) and Diaz and Sanchez (2008) that firms with higher market shares demonstrate market power and are technically efficient compared to firms with lower market shares.Moreover, classifications show a positive effect on technical inefficiency and the coefficient is significant.This finding suggests that basic and consumer classifications are technically inefficient than miscellaneous type.Lastly, time has a positive effect: an indication that technical inefficiency is present in production over time.This Prabowo and Cabanda: Indonesian Firm Efficiency

) 1 (
xk vector of values of known functions of inputs of production and other explanatory variables associated with the i-th firm at the t-th observation; β is a ) 1 ( xk vector of unknown parameters to be estimated; it V s are assumed to be iid ) non-negative random variables, associated with technical inefficiency of production, which assumed to be independently distributed, such that it U is obtained by truncation (at zero) of the normal distribution with mean 

FU
represents the natural log of inventory ( it I ) x the natural log of capital ( it K ); 8  represents the natural log of inventory ( it I ) x the natural log of labor ( it L );Prabowo and Cabanda: Indonesian Firm Efficiency  represents the natural log of fixed assets ( it are non-negative random variable.
model, Equation (2) presents the empirical log-linear form for this research:

Table 3 :
The maximum-likelihood estimates of parameters of the translog stochastic frontier production function for sales a significant positive effect (0.7182) on technical efficiency.The positive effect implies that the manufacturing sector firms' efficiency increases as more inventory utilized.