Los PPA como producto noticioso

Assuming a Cobb-Douglas technology, the production function to be estimated is given below as

where Yt , the output variable is measured using value added data; Xn is the number of employed workers; Xi2 is capital; and n is the number of firms in the industry.2 Moreover, ß0(the intercept term for the ith firm is equal to ß0 + e f and ß y represents the actual response of the output of firm i to the method of application of input j. Finally, s ■ refers to white noise.

The unit of analysis in this study is the manufacturing establishments, classified at the four digit Philippine Standard Industrial Classification (PSIC). This study employs 79 four-digit industries to estimate technical efficiency in 10 manufacturing industries at the three digit level in the Philippines, observed from 1985 to 1988. The four to five digit level industries approach has often been used in the empirical literature to compute for industry technical efficiency due to the unavailability of firm level data3 (Torii 1992; Yoo 1992; Harris 1992; Torii and Caves 1992). A summary of the list of industries included in this study is given in Table 6.1. Only three digit industries which have at least 5 corresponding 4-digit sub-categories are included in the study, each observed in four time periods. The minimum requirement of at least 5

2 The applicability of the Cobb Douglas Production Function was tested using the Ramsey RESET (Regression Specification Error Test) Test. This comes as an option in the autoreg

procedure in SAS. The results indicate that at the 5 per cent level of significance, only one (food industry in 1987) of the forty estimated equations had a misspecification problem. Since the use of the Cobb Douglas form of technology is valid for most of the industries, it was the production function that was used to estimate technical efficiency.

3Firm level data are often not available due to confidentiality requirements as in case of Australia, the Philippines and in many other countries.

(6.13)

j = i

sub-industries is imposed by TERAN, the program used to compute technical efficiency (Kalirajan and Obwona 1994).

Table 6.1 Philippines: list of industries covered in this study, 1985-88

Industry PSIC Number of 4-digit PSIC

Industries Food 311,312 17 Textiles 321 7 Wood 331 7 Furniture 332 5 Non-Metallic 361,362,363, 369 6 Chemical Industries 351, 352 8

Metals and Metal Products 371,372,381 13

Machinery 382 5

Electrical Machinery 383 6

Transport Equipment 384 5

Source: the Author's.

The data is drawn from the annual survey of large manufacturing establishments in the Philippines, undertaken by the National Statistics Office. For consistency, only firms employing 20 or more workers are considered. Output is measured as value added at current prices deflated by implicit price indices based on 1985 prices. Labor input in this model is represented by the total number of paid employees in each industry. A capital stock series at constant prices for each individual industry was constructed as a proxy for capital services in the model. The data were constructed using the perpetual inventory approach (Austria 1992; Hooley 1985).

The NSO survey publishes two statistics which can be used in computing the capital stock data: fixed assets and investment expenditure. Some studies estimate capital stock simply by adjusting the depreciated values of fixed assets. This poses a major problem because since 1970, it has been a standard accounting procedure to value fixed assets at replacement cost, not at original cost. Estimates of which reflect the subjective evaluations so fixed asset values are not standard across firms. Even the date of reappraisals is not uniform so it is not clear whether the assets are being expressed at replacement

cost of year t, or year t-1 (Hooley 1985). To overcome these problems, a capital stock series was constructed from the investment expenditure data at constant 1972 prices, using the perpetual inventory method,

where Kt is the capital stock, 6 is the depreciation rate and /, is the investment expenditure in each period for each industry. The depreciation rate varies between industries and is obtained from the survey based on the ratio of depreciation cost to the book value of asset. The average depreciation rate is

10 per cent.

Equation 6.14 requires an estimate of the initial capital stock. This is done using the formula below where y is the estimated growth rate and 5 is the depreciation rate.

On the basis of the Breusch-Pagan LM test, the null hypothesis, HoiT' = 0 was rejected at the 5 per cent level in all the 10 industries; thus, lending support to the use of the random coefficient regression model (Appendix 6.1). Estimation was by done using the 4-digit level PSIC in each industry, in each time period. The parameters were estimated using weighted least squares to take into account the presence of heteroscedasticity.

Technical efficiency estimates in this study are given in percentages. A technical efficiency estimate of 70.9, for example, suggests that the industry is around 71 per cent efficient relative to its potential output, which is based on the coefficients of the 'best practice technique' in the sample. Overall, estimates generated in this study show wide variability across industries over

Kt = (1 - 5 )Kt_x + I t (6.14)

(6.15)

time (see Figure 6.1 and Appendix 6.2). The results indicate that none of the manufacturing industry in the Philippines is technically efficient. Grouping these industries by end-use classification indicates that Philippine manufacturing is most inefficient in the capital good sector.4 The mean technical efficiency estimate of the consumer good sector represented by the food and furniture industry is 63 per cent over a four year period. For texile, wood, non-metallic and chemical industries which could be considered as intermediate good industries, the mean technical efficiency over the four year period is 72 per cent. However, the capital goods sector which is composed of metal products, machinery, electrical machinery and transport equipment registered the lowest mean technical efficiency estimate of 45 percent over a four year period. The capital good sector being relatively inefficient raise concern about the country’s long run development since this sector affect other downstream industries that uses its output. As Tecson (1996) noted, the structural weakness of the Philippines’ manufacturing sector as shown by the shrinking intermediate and capital goods sector is largely responsible for the country’s high degree of import dependence to run its industries.

4 Manufacturing industries by end-usage is classified as consumer, intermediate or capital goods. Consumer goods include food beverage, tobacco, apparel, footwear and furnitures. The intermediate goods sector on the other hand is consist of textile, leather, wood, paper, printing, chemicals and non-metallic products. Finally, capital goods include metals, machinery, transport equipment and professional equipment (Pineda 1997;Tecson 1996).

Figure 6.1

Technical Efficiency Estimates

□ 1985 □ 1986 □ 1987 ■ 1988 50 -- 20 - - 10 -

Machinery Electrical Transport Machinery Equipment

Source: Author's estimation based on random coefficient approach.

Some caveats regarding the technical efficiency estimates. While the random coefficient approach facilitated the identification of a benchmark potential output in a given sample, comparison of performance are made in relation to the dominant observation in the sample; thus, an industry may be inefficient but since most firms are closer to the degree of efficiency of the ‘best practice’ firm, the industry’s technical efficiency may be high. This explains why textile industry in the study despite being documented as an inefficient industry in the Philippines (Pack 1987) appear to be more efficient than the electronic industry which include electronics and is one of the biggest exports of the country. Second, as indicated by Stevenson (1980) and Caves (1990) different assumptions about firm-specific effects make comparisons from different studies on technical efficiency less meaningful. For example in the case of stochastic frontier approach, imposing the assumption of constant returns to scale may cause the mean response coefficients to be intractable. This is because even when the condition of constant returns to scale is imposed on the response coefficients ( ß y ’s ), the possibility that ^ ß *)1 can not be ruled out due to the relationship that ß * = |max ß + v } (Kalirajan

and Shand 1994). Other authors (Kalaitzandonakes et al 1992, Gong and Sickels 1992) who tried to compare results of technical efficiency estimates using various approaches found that efficiency varied widely when the data envelopment approach (DEA) and the stochastic frontier approach were used in the estimation procedures. Moreover Button and Weyman-Jones (1992) showed that in many cases the stochastic frontier approach and the DEA not only yielded different estimates but also provided different distributions of efficiencies among observations for the same data set. Based on all these comparative studies, Kalirajan and Shand (1994) suggest that efficiency measurement is determined by the choice of functional forms considered to represent the production technology. The data envelopment approach of measuring technical efficiency may be appropriate if the underlying technology is generally weak while information on scale and substitution possibilities is best handled by the stochastic frontier production function approach.