Impact of regulatory reforms on labour efficiency in the Indian and Pakistani commercial banks.

Jaffry, Shabbar ; Ghulam, Yaseen ; Cox, Joe 等

1. INTRODUCTION

Efficiency plays an important role in the operation of firms. If firms are pursing a policy of shareholder wealth maximisation, this implies that maximum efficiency is extracted from a firm's resources during the production process, or that the minimum quantity of inputs are used to achieve a desired level of output. This is especially true in the case of labour demand and labour usage, as wage expenditure constitutes a significant portion of the average firm's cost structure.

Knowledge of relative inefficiencies in labour usage will therefore be of great interest to firm and, as such, academic studies on efficiency of labour demand in firms have been relatively forthcoming. These include work on the Indian farming industry [Kumbhakar (1996), Swedish social insurance offices [Kumbhakar and Hjalmarsson (1991)], Tunisian Manufacturing [Haouras, et al. (2003) and Kalimantanian rice production [Padoch (1985)].

However, there is relatively little in the way of research conducted on efficiency within the banking sector, and even less on the banking sectors of developing economies [Berger and Humphrey (1997)], despite an increase in research activity in such areas over the last ten years. This is unfortunate, as banks and financial institutions are the most important organisations in overall financial intermediation and economic acceleration of a country, in no small part due to their significant role of converting deposits into productive investment. [Podder and Mamun (2004)].

The process of liberalisation and modernisation is vitally important in this particular case. Because of the unique position that it occupies within the framework of an economy, the banking industry tends to be more heavily regulated and scrutinised than other industries. This trend is particularly apparent in developing economies, where banks tend to exhibit poor performance as a result of overly prohibitive regulation [Kumbakhar and Sarkar (2003)]. Thus, tests of labour demand efficiency can be made more meaningful by including some comparison of efficiency both pre and most modernisation. Not only will this paper seek to make comparisons of labour demand efficiency between India and Pakistan, but will also examine changes in the efficiency of labour demand in both the pre and post deregulation periods.

2. CASE STUDIES

A major component of this study will be an examination of the banking sectors of developing economies, and their response to changes in the regulatory environment. In the 1980s and 1990s, a large number of developing economies undertook extensive processes of liberalisation and modernisation, particularly with respect to financial and banking industries. The developed world led the way in this respect, with most notably the USA experiencing productivity and efficiency increases as a result of the relaxation of the country's regulatory environment.

A number of studies have documented this phenomenon within various American industries, including air transportation, telecommunication and freight transportation. Theory does not dictate a clear expected result of deregulation and modernisation in the banking sector in terms of labour efficiency gains (or losses), as the consequences of deregulation may depend on industry conditions prior to the deregulation process, as well as the type of deregulation employed [Berger and Humphrey (1997)].

Bonnaccorsi di Patti and Hardy (2005) examined the efficiency of the Pakistani Banking sector in isolation. Over the period of modernisation, they observe an increase in efficiency as a result of the new competitive environment resulting from the first round of deregulation. It was also found that new private banks sometimes outperformed foreign banks in terms of efficiency.

As a result of this, studies of efficiency in banking, however have not displayed as clear-cut trends as are illustrated in the above examples. Expectations upon the result of the modernisation and deregulation of the banking industries in the countries of the Indian sub-continent are therefore unclear.

2.1. The Indian Banking Sector

India is a country in the heart of the Indian sub-continent of South Asia and has the second largest population in the world. The country was a part of the British Empire until it was recognised as a republic shortly after the end of the Second World War. Owing to its large population India's GDP purchasing power per capita works out to be just US$3,262, ranked 120th in the World by the World Bank. India has shed its dependence on agriculture since it has become a republic. Now, some of the fastest growing industries include IT, textiles and mining.

Banking has also become an emerging industry in the modern era. The Bengal Bank was the first British patronised modern bank in India, and was established in 1784. There are currently a total of 361 different banks in India--27 Nationalised commercial banks, 30 local private banks, 40 foreign commercial banks, 196 regional rural banks and 68 Co-operative banks. The banking sector in India has historically been highly regulated, but gradually the restrictions imposed by such a regime are being lifted. The first 'wave' of reform began in 1969, when fourteen major banks were nationalised. Six more commercial banks were nationalised in 1980. A number of new reforms were introduced during the period 1992-1997. These include a reduction in reserve requirements, privatisation of public sector banks, interest rate deregulation, and an effort to remove barriers to market entry. As such, the Indian banking sector is currently in a transitional phase. Public sector banks are also trying to reduce manpower, non-performing assets and government equity. Foreign direct investment ceilings have also increased under the reforms.

However, since the early 1990s, public sector banks have found it extremely difficult to compete against private sector banks and foreign banks. In response to this, public sector banks are in the process of cutting excessive use of manpower and non-performing assets, as is their right under the new reforms. Other responses to deregulation and modernisation have included an increase in the volume of merger and acquisition activity, an increase in the use of technology, an increased usage of diversification and more sophisticated risk management techniques.

Looking to the future, there has been (and continues to be) an attitude towards the gradual reduction of interest rates by India. GDP growth is expected to continue rising at a rate of around 6 percent per annum. Foreign banks are also likely to meet with more success due to their use of innovative technology and increased freedom afforded by reforms. Indian banks are also expected to move towards a more streamlined and efficient workforce over the coming years.

2.2. The Pakistani Banking Sector

Pakistan also gained independence from the British Empire at the same time as India, with a Pakistani central bank established in 1948. Currently, there are a total of 45 different banks in Pakistan--5 Nationalised commercial banks, 16 local private banks, 19 foreign commercial banks and 5 specialised development banks. Whilst initially the country was very poor, with a significant portion of national wealth generated from agricultural activities, in the modern era the country growth rat has been consistently above the world average.

Despite impressive rates of growth in GDP in recent years, the country experienced an economic slowdown in the early 1990s as a result of poor policy making where the activities of Pakistani banks were focused around subsidising the fiscal deficit, serving a few large corporations and engaging in trade financing [Husain (2005)]. Additionally the financial system suffered from political interference in lending decisions and also in the appointment of banking managers.

In response to some of these problems, a period of de-regulation and financial liberalisation was implemented in the early and late 1990s. These included strengthening of prudential regulations, a market driven exchange rate system, and the appointment of independent persons to the board of directors of nationalised banks and an enhanced capital adequacy requirement and a reduction in the restriction on branching.

Reform of the banking sector is now entering a second phase, where local banks are being asked to raise their paid capital, follow a maximum disclosure requirement and make full provision against non-performing loans. Foreign banks have thrived in the past due to significant investments in technology, including ATMs and credit cards. However, at the current point in time, many foreign banks are selling to local banks [Kazmi (2002)]. The fall in fortunes of foreign banks can be put down to, in part, an increased confidence in privatised domestic banks.

There are a number of factors that are expected to play a part in the development of the Pakistani banking sector, including a gradual reduction in interest rates, an increase in merger and acquisition activity, banks attempting to enter the market for consumer finance, the introduction of new technology and a reduction in non-performing loans. There has also been a rapid rise in branch networks. In 2005, thanks to the ongoing process of reform, almost 80 percent of the banking sector was in private hands [World Bank Report (2005)].

3. FINDINGS OF OTHER RESEARCH

This paper seeks to examine the labour demand efficiency of the banking sectors in India and Pakistan, using the framework outlined by Heshmati (2002), and used in his analysis of labour efficiency within Swedish savings banks. This study will focus on the time period 1985-2003, which is characterised within the Indian sub-continent as a period of significant reform, deregulation and liberalisation in both countries' respective banking sectors.

Even though this paper adopts Heshmati (2002)'s approach, it is worth noting that here are a number of alternative approaches to the measurement of the efficiency of labour demand. Kumbhakar and Sarkar (2003), for example, use TFP growth as the measure of banking performance over the period 1985-1996, including both labour and capital are the variable inputs, while equity and reserves are a quasi-fixed input. The study finds that there is a significant over employment of labour relative to capital, particularly in the public sector, both pre and post deregulation. In contrast to this, Atkinson and Primon (2002) formulate shadow distance and shadow cost systems using panel data for 43 US utilities over 37 years and diagnose an over-use of capital relative to labour and energy and the under-use of energy relative to labour.

Baltagi and Rich (2004) develop a general index time path for technical change between production and non-production labour in US manufacturing industries between 1959-1996. Their findings confirm that substantial reductions in the relative share of labour in the production process is attributable to a sustained period of non-neutral technical change. The general index approach also explains observed shifts in relative labour demand as a combination of price-induced substitution, output effects and skill-biased technical change responses.

In terms of studies focusing on banking, Gjirja uses a translog stochastic frontier input requirement model to assess efficiency in Swedish Bank's use of labour over the period between 1982 and 1998. The study illustrates how deregulation in Sweden positively affected productivity growth, but had no positive impact on the efficiency of labour use. The study also notes that banks were not able to catch up with the expansion of the labour use frontier over time.

Heshmanti (2002) also studies the Swedish banking sector over the period of deregulation in the 1980s, and banking crisis in 1992, looking at a panel of 52 savings banks. The study concludes that the process (and anticipation) of deregulation had a significant affect upon banks' choices of input and output volumes. The study concludes that very small banks tend to operate with a technically optimal size of labour, as the model output indicates a negative relationship between technical efficiency and the size of banks.

Battese, et al. (2000) also examine Swedish Banks covering the period from 1984 to 1995. The study concludes that inefficiency is positive over this period, and has increased--indicating that the 'average' bank did not manage to catch up with the labour use frontier. The study concludes that Sweden might have expected greater effects from deregulation and crisis on labour efficiency and points to the competitive pressure from abroad remaining weak as one of the possible reasons why this was so.

Estache and Rossi (2004) investigate the impacts of different regulatory environments upon the efficiency of firms. The study indicates that privatised firms operating under price-cap and similar 'hybrid' schemes are more efficient in their use of labour than both public firms and privatised firms under rate-of-return regulations, and that privatised firms operating under rate-of-return regulation have, at most, similar labour efficiency as public firms.

Soderbom and Teal (2004) investigate efficiency within the developing African economies and show that the Cobb-Douglas functional form adequately captures efficiency in production technology. The study also concludes that large firms facing higher relative labour costs than smaller firms use a much more capital intensive technology and operate with costs 20-25 percent higher than those which would occur if factor prices differentials across firms of differing sizes could be eliminated.

There have also been a number of studies conducted analysing firm's response in terms of risk preference in the post-deregulation period. Just and Pope (1978), for example, say that risk adverse producers take into account both the mean and variance of output when ranking different technologies, and that this can have an effect upon relative efficiency levels. Rao (2004) investigates cost efficiencies and its relationship with risk-return behaviour of banks in United Arab Emirates by using Stochastic Frontier Analysis in both translog and flexible Fourier forms. The authors detect substantial inefficiencies. In addition, the study concludes that domestic and large banks were less cost efficient than foreign and small banks. The study also revealed a positive and significant relationship between cost efficiencies and levels of capitalisation

4. METHODOLOGY

The model to be used in this study is from Heshmati (2002), itself adapted from the work outlined in Aigner, et al. (1977). In this paper, Heshmati expresses the function of labour demand as:

h = f([y.sub.j],w,q,t)exp([epsilon])

[epsilon] = [mu] + [upsilon] ... ... ... ... ... ... ... (1)

Where h is units of labour measured in hours, t represents the production technology and yi (j = 1,2,..., M) are services produced using labour, w is wage, q is a vector of quasi-fixed factors and t is time effects. This function estimates the minimum amount of labour required to produce a given level of output. The error term in this equation is decomposed into two distinct parts (u and x>), representing technical efficiency and factors beyond the control of banks respectively. In addition to these two, the bank's production technology will also have an effect upon their demand for labour.

If the [mu] component of the error term is greater than or equal to zero, the firm displays a level of technical inefficiency [Aigner, et al. (1977)], as the firm has used more labour than was technically necessary in order to produce a given level of output. A bank, which displays s a [mu] value of zero, can claim to be fully efficient in the use of labour. The [upsilon] component of the error term can be both positive and negative. Due to its presence, therefore, the labour demand frontier is stochastic even when u is set to zero.

If risk functions are also taken into account, then the model is redefined appropriately. Again, as from Heshmati (2002), Robinson and Barry (1987) and Just and Pope (1978). When doing so, the model then becomes:

h = f(x;[alpha])exp(g(x;[beta])[epsilon]) ... ... ... ... ... ... (2)

Where x = (y,w,q,t), with /(-x;[alpha]) representing the demand part and g(.Y;P)s representing the variance part of the demand function. The model can also be re-specified in log linear form.

ln h = ln f(x;[alpha]) + g(x;[beta])[epsilon] ... ... ... ... ... ... (3)

5. DATA

Panel data is taken from a selection of Indian and Pakistani banks, covering the period 1985-2003. The data were taken from the annual reports of 73 Indian and 41 Pakistani banks, and are each provided at bank levels, rather than at the individual branch level. Any conclusions generated form this study can therefore only be made at a bank level. Unlike Das, et al. (2005), specific branch level analysis is not possible using the data available for this study.

The panel consists of 1 14 Indian and Pakistani commercial banks observed for 19 years (from 1985 to 2003), and is unbalanced as not all banks were in existence for the whole sample period. Summary statistics of the data are presented in Table 1 (below).

Outputs and inputs are chosen as per policy objectives of the individual banks, as well as those of the regulatory reforms within the respective countries. The specific variables used in the analysis include the total quantity of labour hours used (h), wages (w), loans (y1), investment (y2), deposits (y3), number of branches (y4), fixed assets (q), and a time trend (/) representing exogenous rates of technical change. The wages, loans, deposits, investment, fixed assets and total labour costs are provided in constant 2000 prices United States Dollar (USD) values to make the data comparable between India and Pakistan. Labour is measured in hours used per year. As the data on number of hours worked for each employee is not available, we have used a rough proxy figure of 2400 considering 300 working days and eight hours per day work. The 'wage' variable is defined as hourly wages--an aggregate measure of the cost associated with the hiring of labour, including payroll taxes. The quasi-fixed variable, q, is defined as the sum of fixed assets. Quality and risk variables are used in the regression stages to control for heterogeneity in risk taking behaviour.

In this study, the variables loans, investment deposits and branches are regarded as outputs. The literature is divided as to whether certain variables are an input or an output. Berger and Humphry (1997), for example, see deposits as outputs (what is known as a value-added approach). In the case of this study, it is important to define whether total number of branches is considered to be an input or an output. In this case, as with Heshmati (2002) and Kumbhakar and Sarkar (2003), the number of branches is considered an output variable.

It is worth noting at this stage how the 'size' variable was constructed. A size distribution is calculated by the number of employees of each bank, with the following restrictions:

From the total sample of data, 1 percent of observations was determined to be excessively large or small outliers, and was resultantly excluded from the model. The regression outlined below was subsequently run for the combined dataset. In order to make country specific technical efficiency scores apparent, individual bank specific efficiencies are calculated and then separated by years, size classes, types of ownership and countries.

A flexible translog functional form (which is linear in parameters) is then used to approximate/(.). The model can therefore be specified as follows;

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)

where h, y, w and q are variables which are defined above, i is an index of banks ... i(1 ... N), /represents an index of time ... t(1 ... T) and both j and k ... j;k (1 ... M) are indices of outputs. Finally, the exogenous rate of technical change is represented by [[lambda].sub.t].

The key to this Equation is the bank, which performs best in terms of technical efficiency within the sample. We assume that this particular bank is fully efficient (hence the u value for this bank is equal to 0). All other banks in the sample are assumed to be inefficient to a certain degree, the extent of which is determined relative to the single, fully efficient bank.

One of the drawbacks associated with this method is that the 'fully efficient' bank may not always be the best in all of the time periods used in the study. For this reason, following Schmidt and Sickles (1984), a time variant technical inefficiency score is calculated (relative to the banks with best performances in each year) as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

And technical efficiency as

[TINEFF.sub.it] = exp(-[TINEFF.sub.it]) ... ... ... ... ... (6)

Which, as Heshmati (2002) points out, is both bank and time specific. The expectations on the first order coefficients are as with Heshmati (2002), where [[alpha].sub.j] and [[alpha].sub.q] are expected to be positive and [alpha]w negative, which can be interpreted s the elasticity of labour demand with respect to output, quasi-fixed inputs and wages respectively. These expectations are only valid at the normalised data point, with the corresponding elasticities (which are both bank and time specific) for all data points derived respectively as follows.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (9)

And the time specific elasticity of labour with respect to time (the exogenous rate of technical change) is derived as:

[E.sub.t] = [partial derivative]ln[h.sub.it]/[partial derivative]t = ([[lambda].sub.t] - [[lambda].sub.t-1]) ... (10)

6. ESTIMATION AND EXPLANATION

The model used in this study follows the approximation outlined in Heshmati (2002) and detailed above. The model outlined in Equation (4) is firstly used to estimate labour demand function for a sample of Indian and Pakistani commercial banks observed for nineteen-year period between 1985-2003. Estimates of the demand function, (f) and the variance function, (g) regression results are presented in Appendix A. A majority of the bank-specific variables are statistically significant from zero. The variance function, (g). outlined above, was estimated using the weighted non-linear method. The results can be seen in Table 3, below, and in the Appendix A.

Labour Demand Elasticity and Productivity Growth

The elasticities of labour demand with respect to different outputs, wages and fixed assets were calculated for bank and time specific. The mean values are reported in Table 3 (above) by years, bank size, type of ownership and country. At the mean data point, all elasticities with the exception of wage and time elasticities are positive and significant, indicating that there is a degree of responsiveness of labour demand to changes in the levels of outputs and wages and fixed assets. These are similar results to those provided by Heshmati (2002). Labour demand elasticity with respect to the output 'loans' is negative for a number of specific bank sizes--namely small, large, very large and public banks. It is possible that some large or small banks are leaning toward an 'arm's length' approach to banking, perhaps using less labour and more automation in order to generate new loans.

In relative terms, the sample mean elasticities of labour with respect to 'loans' and 'investments' are quite small, taking values of 0.03 I, 0.054 respectively (although these are still slightly higher than Heshmati's estimates of the same elasticities applied to Swedish savings banks), with standard deviations which are not unusually large in either case. Both of these elasticities are steadily decreasing over time, suggesting that loans and investment have become less labour intensive outputs for Indian and Pakistani banks over time. 'Deposits' have larger labour demand elasticity than either of these two outputs, with a sample mean elasticity of 0.175 and a standard deviation of 0.057. This means that 'deposits' are a significantly more labour intensive output than 'loans' and 'investments'.

The largest of the output elasticities is that which relates to 'branches', with a mean elasticity of 0.452 and a standard deviation of 0.009. We can conclude from this that 'branches' is the most labour-intensive banking service offered in India and Pakistan, and this degree of elasticity is very unlikely to be significantly different across the number of banks included in the sample. This is a result that is to be expected, showing that a marginal change in the number of branches will have the largest marginal effect on labour demand of all outputs.

The wage elasticity is on average -0.288 (the largest of the input elasticities) with a relatively small standard deviation of 0.10. In contrast to the results of Heshmati (2002), the elasticity of wages is decreasing over time, from -0.442 in 1985 to -0.163 in 2003. Although the sign of this variable is consistent with theory, such changes in elasticity over time contradict Heshmati (2002) and indicate that labour demand is becoming less and less responsive to changes in the wage rate. This seems to indicate that banking reforms within the Indian sub-continent have not had the desired effect of making labour use more efficient.

The elasticity with respect to fixed assets has a mean value of 0.010 with a relatively high standard deviation of 0.007. There is an upward trend in the elasticity of this input, starting at -0.002 in 1985 and ending at 0.014 in 2003. Again, these results contrast with Heshmati (2002), and indicate an increasing demand for labour with the accumulation of fixed assets. There must be significant differences in the degree to which the crowding out of labour as a result of changes in wage levels in the sub-continent as compared to Sweden.

The exogenous rate of technical change (consisting of only a neutral component) changes only over time. The sample mean value is -0.007, indicating very slight positive technical change (due to a slight reduction in labour usage). Technical change fluctuates from positive to negative before the turn of the 21st Century, with no clear trend established for more than a few tears at a time. However, from 1999, there is a consistent positive technical change. It would be interesting to see if this trend has continued post 2003, and further study might seek to establish whether or not this apparent trend towards positive technical change is consistent throughout the decade. However, when taking the overall sample into account, it appears that there has been no definite trend in terms of technological progress or regress in the sub-continent over the years included in the study.

The magnitude of the different output elasticities also appear to vary with the size of the individual bank in question. Labour demand elasticity with respect to wages seems to fall as bank size increases. This illustrates that increased wage levels do not serve as great a deterrent to the hire of additional labour for larger banks as opposed to their smaller counterparts. The elasticity with respect to branches decreases with size of bank, showing that larger banks are better able to expand their branching network without having the large effect on labour demand experienced by smaller banks across the sub-continent. Labour demand elasticity with respect to fixed assets seems to fall very slightly with the size of the bank in question (showing that increasing volumes of fixed assets requires less additional labour for larger banks). Finally, the time trend shows very consistent amounts of technical change for all sizes of bank, meaning that the very small technical progress shown over the sample period has not been limited to banks of specific sizes. Aside from these, there does not appear to be a definite trend with respect to the other input or output labour demand elasticities as bank size increases.

In terms of elasticity differentiated by bank ownership, public banks have negative labour demand elasticities with respect to loans, while foreign banks have by far the largest elasticity with respect to this output. It seems that public banks in the sub-continent are leading the way in reducing the labour intensity associated with marginal increases in the production of this output. Other differentials of note include foreign banks having significantly more labour demand elasticity with respect to branches; while public banks have the lowest. This means that it is publicly owned banks that are able to expand their branch network with the smallest marginal impact on labour demand. This ,nay be due to more efficient management and organization, or to economies of scale. Public banks also have the largest elasticity with respect to wages, with foreign banks having almost half the labour demand elasticity with respect to wages than their public counterparts. Private, domestic banks are fairly close to public banks with regard to elasticities applicable to this particular input. This means that domestic banks in the sub-continent are far more responsive in their demand for labour when the wage rate changes than their foreign competitors.

The Employment Variance

The figures representing the following can be found in Appendix A (Section B). The beta coefficients with respect to 'investment' (y2), 'deposits' (y3) and 'branches' (y4) are all positive and statistically significant. The coefficient for 'loans' (y1) is however negative but statistically insignificant at the 5 percent level of significance. Of all of the input variables, the wage coefficient is by far the largest, and, as with Heshmati (2002) is both negative and strongly significant. Time specific dummy variables have a mixed signs and few of them statistically significant. The coefficient applying to fixed assets is positive, but not significant.

In common with Heshmati (2002), the employment variance elasticity or marginal risk effects are calculated with respect to the dispersion factors of 'Outputs', 'Wages', 'Fixed assets' and 'Time Trend', with mean values being estimated separately for each year, size of bank, type of bank and country. These results, together with the overall sample mean, are reported in second part of Table 3. Marginal variance (risk) effects evaluated at the mean of the data with respect to 'loans', 'investment', 'deposits' and 'branches' are generally negative. Positive marginal effects are observed for wages in post deregulation period. In all cases the standard deviations are large and, for some variables, are in excess of the mean value itself. Thus, generally for banks with production levels close to the sample means, the employment variance decreases if the bank produces more output.

The variables 'Wages' and 'Branches' are the most important factors contributing to the variance of employment in terms of marginal effects. The signs of marginal effects are, on the whole, as expected. Significantly more variation in the estimated marginal effects seems to take place almost uniformly as bank size increases. Some of the inputs and outputs seem to demonstrate significant variation in their respective marginal effects over time (notably loans, branches and wages), while the remaining inputs and outputs display fairly consistent marginal effects over time.

Technical Efficiency

The efficiency measured here is a relative efficiency, as it is measured relative to the bank demonstrating 'best-practice' in each year. This individual bank is assumed to be 100 percent efficient. The mean values of estimates of technical efficiency obtained from Equation (6) are reported in Table 3 by year, bank size, type of bank and country. Technical efficiency is both bank and time-specific. The overall mean technical efficiency is 65.3 percent with a standard deviation 0.132. This means that, on average, banks in the sub-continent could have reduced their labour usage by 34.7 percent with output remaining constant. This is indicative of a relatively low level of mean labour use efficiency displayed by banks in the sub-continent over the sample period.

However, what is apparent from investigating the changes in labour use efficiency over time is that the financial reforms initiated in the 1990s have helped to improve the efficiency of labour demand within banks, as the mean technical efficiency over time is increasing. In 1985, the average commercials bank in the sub-continent showed only 40 percent labour use efficiency, compared with 65 percent in the final year of the sample period, and a high of 82 percent in 1995. The year on year change is largest between 1990 and 1992 (despite a small adjustment 'blip in 1991), which is indicative of the success of the round of reforms introduced in 1992. There appears to be a noticeable variation in technical efficiency over the bank size. As was concluded in the study of Heshmati (2002), there is found to be a negative relationship between the level of technical efficiency and the size of banks. In a relative sense, very small banks operate with a more technically optimal size of labour than do very large banks. The results indicate that the largest banks could reduce their labour demand on average by 43.6 percent. Therefore, there is a very significant gap between the optimal level of labour efficiency, and that, which is observed in the largest banks within the sub-continent. The very smallest sub-continental banks were found to be slightly inefficient in labour usage, and could have reduced labour usage by 25.9 percent.

Among banks of different ownership types, it was found that foreign banks were the most efficient in terms of labour usage, followed by private domestic and public domestic commercials banks respectively. Foreign banks could have reduced their labour usage by 26.1 percent, private banks by 33.5 percent and public banks by 42 percent, indicating that publicly owned commercials banks in the sub-continent are employing far more labour than is technically necessary given output levels, and still have some way to go in improving technical efficiency levels in the future. The frequency distribution of technical efficiency is reported in Table 4. A significant number of banks are found in the intervals of between 60 percent and 80 percent labour usage efficiency.

The correlation coefficients of ranking of efficiencies are reported in Table 5. This study concludes that there is a negative correlation between efficiency and the size of bank, while a positive relationship is found between efficiency and time. We find a positive association between both the input and output variance effects and time, which is significant for the former, but not for the latter. We also find a negative relationship between the input and output variance effects and bank size where input and output variance effects refer to the sum of total marginal effects with respect to input and output variables.

7. CONCLUSION

This paper has sought to examine the efficiency of labour use in both India and Pakistan during a period of modernisation and deregulation. Data from 73 Indian and 41 Pakistani banks have been analysed over the period 1985-2003. A flexible translog functional form is used where demand for labour is a function of wages, fixed inputs and a time trend. Of those outputs and inputs elasticities are largely as expected. The largest elasticity is with respect to wages, which have a strong negative elasticity. Of outputs, branches have the most effect upon labour demand, with a strong, positive elasticity.

The most interesting conclusions from this study are those that illustrate technical efficiency levels. The average level of technical inefficiency across the sample was relatively low, as was expected. It was found that, on average, banks in the sub-continent could have reduced their labour usage by 34.7 percent with output remaining constant. However, the sub-continent was generally experiencing increases in labour efficiency across the nineteen years of the study, indicating that policies enacted in the early and late 1990s to assist banks in the reduction of their labour use were reasonably successful. This level of efficiency varies inversely with bank size as expected. The results indicate that the largest commercial banks could reduce their labour demand on average by 43.6 percent. The very smallest banks were found to be slightly more efficient on average in terms of labour usage, which could have been reduced by 25.9 percent. Among banks of different ownership types, it was found that foreign banks were the most efficient in terms of labour usage, followed by private domestic and public domestic commercial banks respectively.

It would appear that the significant financial reforms of the last decade in the Indian subcontinent over the last decade have reduced the degree of over-usage of labour in its banking sectors. There still exists, however, a fairly large degree of inefficiency in terms of labour usage, particularly among the very large banks of the subcontinent. It appears that the number of branches that are owned by a bank have the greatest impact on the demand for labour and, if the outcome of more efficient labour usage is to be achieved, more emphasis needs to be placed on those large banks with an extensive network of branches. These are most likely to be the banks that have previously been publicly owned, and therefore may have encountered difficulties meeting the challenges of the new competitive environment. It may be deemed that additional effort needs to be made to streamline these large banks if the desired efficiency gains are to be made.

REFERENCES

Aignes, D. J., C. A. K. Lovell, and P. Schmidt (1977) Formulation and Estimation of Stochastic Production Function Models. Journal of Econometrics 6, 21-37.

Atkinson, S. E., and D. Primon (2002) Stochastic Estimation of Firm Technology, Ineffiency and Productivity Growth Using Shadow Cost and Distance Functions. Journal of Econometrics 108, 203-225.

Baltagi, B. H., and D. P. Rich (2005) Skill Based Technical Change in US Manufacturing: A General Index Approach. Journal of Econometrics 126, 549-570.

Battese, G. E., A. Heshmanti, and L. Hjalmarsson (2000) Efficiency of Labour Use in the Swedish Banking Industry: A Stochastic Frontier Approach. Empirical Economics 25, 623-640.

Berger, A. N. and D. B. Humphrey (1997) Efficiency of Financial Institutions: International Survey and Directions for Future Research. European Journal of Operational Research 98, 175-212.

Bonnaccorsi di Patti, B. and D. C. Hardy (2005) Financial Sector Liberalisation, Bank Privatisation and Efficiency: Evidence from Pakistan. Journal of Banking and Finance 29, 2381- 2406.

Das, A., Ray, S. C. and A. Nag (2005) Labour Use Efficiency in Indian Banking--A Branch Level Analysis. University of Connecticut, Department of Economics, Working Paper Series.

Estache, A. and M. A. Rossi (2005) Do Regulation and Ownership Drive the Efficiency of Electricity Distribution? Evidence from Latin America. Economics Letters 86: 2, 253-257.

Gjirja (n.d.) Effects of Deregulation and Crisis on Labour Use Efficiency in the Swedish Banking Industry. PhD thesis, Gtiteborg University.

Haouras, I., M. Yagoubi, and A. Heshmati (2003) Labour Use Efficiency in Tunisian Manufacturing Industries. Review of Middle-East Economics and Finance 1, 3, 195-214.

Heshmati, A. (2002) Labour Demand and Efficiency in Swedish Savings Banks. Applied Financial Economics 11,423-433.

Just, R. E., and R. D. Pope (1978) Stochastic Specification of Production Functions and Economic Implications. Journal of Econometrics 7: 67-86.

Kumbhakar, S. C., and L. Hjalmrasson (1991) Labour Use Efficiency in Swedish Social Insurance Offices. Institute for the Study of Labour (IZA). (IZA Discussion Papers, 687.)

Kumbhakar, S. C. (1996) A Farm-level Study of Labour Use and Efficiency Wages in Indian Agriculture. Journal of Econometrics 72:177-195.

Kumbhakar, S. C. and S. Sarkar (2003) Deregulation, Ownership and Productivity Growth in the Banking Industry: Evidence from India. Journal of Money, Credit and Banking 35: 3,403-424.

Padoch, C. (1985) Labour Efficiency and Intensity of Land Use in Rice Production: An Example from Kalimantan. Human Ecology 13:1 3,271-289.

Podder, J. and A. A. Mamun (2004) Loan Loss Provisioning System in Bangladesh Banking: A Critical Analysis. Managerial Auditing Journal 19:6, 729-740.

Rao, A. (2005) Cost Frontier Efficiency and Risk-return Analysis in an Emerging Market. International Review of Financial Analysis 14: 3,283-303.

Soderbom, M., and F. Teal (2004) Size and Efficiency in African Manufacturing Firms: Evidence from Firm-level Panel Data. Journal of Development Economies 73, 369-394.

Shabbar Jaffry <shabbar.jaflfryfa/port.ac.uk> is Reader in Economics, Yasccn Ghulam <yasecn. [email protected]> is Director of Postgraduate Programme in Economics, and Joe Cox <[email protected]> is Lecturer. Department of Economics. Portsmouth Business School, University of Portsmouth, UK.

Appendix A

GLS Parameter Estimates of the Labour Demand and Nonlinear Least Square
Estimates of the 0ariance Function (Combined)

A. Labour Demand Function

a0 1.8769 * 0.1418
ay1 -0.0241 0.0315
ay2 0.0901 ** 0.0232
ay3 0.0612 * O.0334
ay4 0.4912 ** 0.0344
aw -0.1204 ** 0.0441
aq 0.0083 0.0290
ay11 0.0269 0.0101
ay32 0.0149 ** 0.0236
ay33 0.0278 0.0033
ay44 0.0121 0.0235
aww 0.0744 ** 0.0066
aqq -0.0003 0.0175
ayl2 -0.0211 0.0031
ayl3 0.0556 0.0184
ay14 -.0316 * 0.0428
ay1w -.0533 0.0141
aylq -.0706 ** 0.0312
ay23 -0.0803 ** 0.0141
ay24 0.0512 ** 0.0219
ay2w 0.1384 ** 0.0098
ay2q 0.0111 0.0195
ay34 -0.0812 ** 0.0087
ey3w -0.2123 ** 0.0183
ay3q 0.0555 ** 0.0409
ay4w 0.1189 ** 0.0176
ay4q 0.0014 0.0137
awq 0.0290 ** 0.0046
C2 0.0127 0.0116
C3 0.0088 0.0154
C4 0.0508 ** 0.0162
CS 0.0213 0.0169
C5 0.0198 0.0177
C7 -0.0002 0.0185
C8 -0.0042 0.0235
C9 -0.0205 0.0245
C10 -0.0197 0.0264
C11 -0.0009 0.0262
C12 -0.0108 0.0270
C13 -0.0367 0.0282
C14 -0.0233 0.0294
C15 -0.0369 0.0307
C16 -0.0858 ** 0.0309
C17 -0.1330 ** 0.0310
C18 -0.1497 ** 0.0319
C19 -0.1426 ** 0.0331
D2 -1.4880 ** 0.0329
D3 -1.4838 ** 0.1400
D4 -1.4838 ** 0.1470
D5 -1.3719 ** 0.1417
D6 -1.4947 ** 0.1538
D7 -1.5394 ** 0.1685
D8 -1.7031 ** 0.1687
D9 -0.9720 ** 0.0836
D10 -0.8144 ** 0.0667
D11 -0.8615 ** 0.0840

d12 -0.8053 ** 0.0885
d13 -0.9500 ** 0.0725
d14 -1.1713 ** 0.0957
d15 -1.2362 ** 0.1120
d16 -1.1402 ** 0.1131
d17 -0.9401 ** 0.1039
d18 -1.0788 ** 0.0990
d19 -1.4766 ** 0.1016
d20 -1.2957 ** 0.1155
d21 -1.5890 ** 0.1455
d22 -1.6336 ** 0.1557
d23 -1.4813 ** 0.1374
d24 -1.5322 ** 0.1220
d25 -1.4590 ** 0.1212
d26 -1.4947 ** 0.1284
d27 -.5407 ** 0.1410
d28 -.9401 ** 0.1717
d29 -1.9097 ** 0.1681
d30 -2.0338 ** 0.1728
d31 -1.9453 ** 0.1749
d32 -2.1090 ** 0.1715
d33 -2.0121 ** 0.1874
d34 -2.2942 ** 0.1908
d35 -2.1664 ** 0.1749
d36 -2.2752 ** 0.1787
d37 -2.1646 ** 0.1799
d38 -2.3285 ** 0.1983
d39 -2.4739 ** 0.1895
d40 -2.1861 ** 0.1903
d41 -2.3707 ** 0.2170
d42 -2.6330 ** 0.1982
d43 -2.5258 ** 0.2081
d44 -2.3914 ** 0.2050
d45 -2.3704 ** 0.1959
d46 -2.8769 ** 0.2137
d47 -2.3685 ** 0.2146
d48 -2.6190 ** 0.2175
d49 -2.8671 ** 0.2154
d50 -3.1282 ** 0.2481
d51 -1.3382 ** 0.2422
d52 -1.3213 ** 0.2179
d53 -1.2972 ** 0.2309
d54 -1.3505 ** 0.2452
d55 -1.7387 ** 0.2474
d56 -1.2204 ** 0.2286
d57 -1.9500 ** 0.2408
d58 -1.9285 ** 0.2399
d59 -2.3957 ** 0.2408
d60 -2.1982 ** 0.2411
d61 -2.6138 ** 0.2419
d62 -2.9448 ** 0.2419
d63 -3.0255 ** 0.2433
d64 -1.9438 ** 0.2433
d65 -2.9248 ** 0.2398
d66 -2.9842 ** 0.2444
d67 -2.9127 ** 0.2404

d68 -2.8301 ** 0.2421
d69 -3.0724 ** 0.2429
d70 -2.8792 ** 0.2429
d71 -2.0926 ** 0.2493
d72 -3.0830 ** 0.2403
d73 -2.4219 ** 0.2375
d74 -1.9263 ** 0.1396
d75 -1.8993 ** 0.2411
d76 -2.1077 ** 0.2357
d77 -2.2246 ** 0.2312
d78 -1.7513 ** 0.2366
d79 -2.7398 ** 0.2297
d80 -2.5103 ** 0.2401
d81 -2.0333 ** 0.2449
d82 -3.0176 ** 0.2299
d83 -.8897 ** 0.2664
d84 -.0300 ** 0.1887
d85 -2.6177 ** 0.2400
d86 -2.2349 ** 0.2331
d87 -2.6660 ** 0.2469
d88 -1.8077 ** 0.2421
d89 -2.2207 ** 0.2429
d90 -2.9928 ** 0.2360
d91 -2.3572 ** 0.2374
d92 -2.4372 ** 0.2388
d93 -2.7708 ** 0.2285
d94 -2.4817 ** 0.2425
d95 -1.2751 ** 0.1003
d96 -2.4669 ** 0.2393
d97 -3.1033 ** 0.2462
d98 -2.8417 ** 0.2385
d99 -1.7112 ** 0.2381
d100 -1.4452 ** 0.1179
d101 -2.7886 ** 0.1106
d102 -2.6981 ** 0.2345
d103 -2.4645 ** 0.2393
d104 -2.7785 ** 0.2339
d105 -3.0457 ** 0.2387
d106 -2.8658 ** 0.2368
d107 -2.6618 ** 0.2443
d108 -1.9528 ** 0.2367
d109 -1.9528 ** 0.2382
d110 -2.8477 ** 0.2476
d111 -2.2803 ** 0.2309
d112 -1.4467 ** 0.1092
Quality -0.0089 * 0.0046
Risk -0.0417 0.0298

B. Variance Function

by1 0.0025 0.0035
by2 0.0006 0.0028
by3 -0.0049 0.0051
by4 -0.0036 0.0037
bw 0.0552 ** 0.0039
bq 0.0029 0.0030
bt 0.0050 ** 0.0003
[[sigma].sub.2v] 5.7433

Table 1
Summary Statistics of the India and Pakistani (Combined)
Commercial Banks, 1985-2003 (in rea1 2000 US dollars)

Variable Variable Name N Mean Std Dev

Idnr Bank ID 1681 51.45 32.88
Period Year 1681 1994 5
Lcost (c) Labour Cost (mill.) 1681 103.39 288.34
Hours (h) Labour Hours (mill.) 1681 26.29 62.93
Wage (w) Hourly wage rate 1681 5.60 5.31
Fixass (q) Fixed assets (mill.) 1681 46.09 89.27
Loans (y1) Loans (mill.) 1681 2582.83 7287.54
Inv (y2) Investment (mill.) 1681 2270.79 5562.63
Deposits y3) Deposits (mill.) 1681 4592.03 11331.70
Brans (y4) Branches 1681 595 1134
T (t) Time trend 1681 10.32 5.34
Size Size of the bank 1681 2.42 1.52
Type Type of bank 1681 1.98 0.81
Qty Quality 1681 6.30 3.28
Lar Risk 1681 42.82 10.57

Variable Variable Name N Minimum Maximum

Idnr Bank ID 1681 1 114
Period Year 1681 1985 2003
Lcost (c) Labour Cost (mill.) 1681 0.07 4382.31
Hours (h) Labour Hours (mill.) 1681 0.03 575.16
Wage (w) Hourly wage rate 1681 0.71 57.72
Fixass (q) Fixed assets (mill.) 1681 0.00 944.50
Loans (y1) Loans (mill.) 1681 0.64 101113.76
Inv (y2) Investment (mill.) 1681 0.00 63210.30
Deposits y3) Deposits (mill.) 1681 3.17 155534.23
Brans (y4) Branches 1681 1 9089
T (t) Time trend 1681 1 19
Size Size of the bank 1681 1 5
Type Type of bank 1681 1 3
Qty Quality 1681 0.01 14.94
Lar Risk 1681 4.91 82.27

Notes: Type of bank includes three categories: public,
private, foreign.

Quality = capital to asset ratio = (capital + reserves)/total
assets) * 100.

Risk = loans to assets ratio = (Total loans/Total assets) *
100.

Type of banks includes three categories: public, private, foreign.

Size of banks includes five categories: very small, small, medium,
large and very large based on number of employees.

Labour hours = labour * 2400.

Table 2
Construction of 'Size' Variable

 Resultant 'Size'
Number of Bank Employees Classification

Employees [less than or equal to] 1,000 1
1,000 < Employees [less than or equal 2
 to] 5,000
5,000 < Employees [less than or equal 3
 to] 10,000
10,000 < Employees [less than or equal 4
 to] 20,000
Employees > 20,000 5

Table 3
Mean Input and Output Elasticities and Marginal Effects by Year,
Size and Type of Banks (Indian and Pakistani (Combined)
Commercial Banks)

 Output

 Loans Investment Deposits
 (y1) (y2) (y3)

A. Labour Demand
Elasticity
 1985 0.058 0.119 0.094
 1936 0.054 0.111 0.104
 1987 0.058 0.113 0.091
 1938 0.049 0.122 0.092
 1989 0.058 0.096 0.119
 1990 0.059 0.084 0.133
 1991 0.048 0.059 0.144
 1992 0.042 0.054 0.148
 1993 0.037 0.036 0.172
 1994 0.035 0.039 0.180
 1995 0.021 0.034 0.212
 1996 0.019 0.021 0.224
 1997 0.010 0.014 0.247
 1998 0.003 0.019 0.237
 1999 0.003 0.019 0.242
 2000 0.011 0.019 0.238
 2001 0.010 0.021 0.228
 2002 0.005 0.024 0.209
 2003 0.010 0.026 0.203
 Very Small 0.085 0.035 0.184
 Small -0.001 0.068 0.169
 Medium 0.001 0.054 0.179
 Large -0.003 0.070 0.149
 Very Large -0.037 0.054 0.209
 Public -0.014 0.060 0.175
 Private 0.016 0.047 0.186
 Foreign 0.091 0.047 0.176
 India 0.014 0.059 0.163
 Pakistan 0.065 0.035 0.214
 Sample Mean 0.031 0.054 0.175
 St. deviation 0.022 0.040 0.057

B. Marginal Variance
(Risk) Effects
 1985 -0.051 -0.164 0.015
 1986 -0.036 -0.001 0.048
 1987 -0.037 0.478 0.011
 1988 -0.066 -0.030 0.013
 1989 -0.027 -0.042 -0.012
 1990 -0.030 -0.041 -0.012
 1991 -0.004 -0.012 -0.001
 1992 -0.004 -0.016 0.005
 1993 -0.003 0.178 -0.002
 1994 -0.896 -0.008 -0.005
 1995 0.003 -0.036 0.002
 1996 1.536 -0.010 -0.003
 1997 -0.003 -0.001 -0.014
 1998 0.002 -0.002 0.059
 1999 0.005 0.000 -0.008
 2000 0.006 -0.009 0.414
 2001 -0.010 -0.001 0.025

 2002 0.078 -0.011 0.024
 2003 0.068 0.003 0.019
 Very Small 0.000 0.000 0.000
 Small -0.002 -0.097 -0.005
 Medium -0.001 0.006 0.019
 Large 0.615 0.028 0.013
 Very Large -0.380 -0.116 0.162
 Public 0.110 -0.032 0.081
 Private 0.000 -0.005 0.001
 Foreign -0.001 -0.060 -0.003
 India -0.085 0.052 0.018
 Pakistan 0.314 0.013 0.048
 Sample Mean 0.028 -0.036 0.026
 St. deviation 0.420 0.122 0.096

 Inputs

 Branches Wages Fixed Asst
 (y4) (w) (q)

A. Labour Demand
Elasticity
 1985 0.443 -0.442 -0.002
 1936 0.444 -0.423 0.000
 1987 0.456 -0.419 0.004
 1938 0.462 -0.419 0.001
 1989 0.450 -0.411 0.003
 1990 0.450 -0.406 0.005
 1991 0.468 -0.298 0.014
 1992 0.464 -0.275 0.019
 1993 0.468 -0.256 0.026
 1994 0.465 -0.269 0.019
 1995 0.453 -0.263 0.010
 1996 0.450 -0.239 0.015
 1997 0.444 -0.233 0.012
 1998 0.452 -0.197 0.014
 1999 0.441 -0.202 0.010
 2000 0.443 -0.213 0.008
 2001 0.443 -0.197 0.010
 2002 0.448 -0.144 0.014
 2003 0.444 -0.163 0.014
 Very Small 0.527 -0.423 0.013
 Small 0.456 -0.215 0.013
 Medium 0.030 -0.175 0.008
 Large 0.385 -0.156 0.008
 Very Large 0.329 -0.155 0.006
 Public 0.365 -0.157 0.009
 Private 0.455 -0.197 0.010
 Foreign 0.541 -0.504 0.014
 India 0.440 -0.223 0.014
 Pakistan 0.479 -0.415 0.004
 Sample Mean 0.452 -0.288 0.010
 St. deviation 0.009 0.100 0.007

B. Marginal Variance
(Risk) Effects
 1985 -0.209 -0.837 0.051
 1986 -0.144 -0.928 -0.039
 1987 -0.084 -0.992 -0.067
 1988 0.266 -1.076 -0.061
 1989 -0.107 -0.162 0.034
 1990 -0.132 0.404 0.022
 1991 -0.031 0.044 -0.010
 1992 0.046 0.281 -0.011
 1993 -0.016 0.046 -0.004
 1994 -0.027 0.004 -0.008
 1995 0.022 0.116 -0.008
 1996 -0.014 0.029 -0.004
 1997 0.009 0.018 -0.004
 1998 -0.010 0.015 -0.004
 1999 0.069 0.028 -0.003
 2000 -0.042 0.097 -0.006
 2001 -0.033 0.040 -0.004
 2002 -0.027 0.061 -0.005
 2003 -0.039 0.087 -0.007
 Very Small 0.000 0.007 0.000
 Small 0.004 -0.005 0.001
 Medium -0.002 -0.050 0.000
 Large -0.762 -0.321 -0.008
 Very Large -0.251 -0.444 -0.104
 Public -0.457 -0.367 -0.050
 Private -0.001 0.016 0.000
 Foreign 0.002 0.009 0.000
 India -0.026 -0.096 -0.024
 Pakistan 0.039 -0.141 -0.002
 Sample Mean -0.168 -0.143 -0.019
 St. deviation 0.512 0.449 0.021

 Time
 (t) TME Efficiency

A. Labour Demand
Elasticity
 1985 -0.000 -- --
 1936 -0.004 -- --
 1987 -0.005 -- --
 1938 -0.054 -- --
 1989 -0.044 -- --
 1990 -0.002 -- --
 1991 -0.041 -- --
 1992 -0.004 -- --
 1993 -0.026 -- --
 1994 -0.005 -- --
 1995 -0.027 -- --
 1996 -0.007 -- --
 1997 -0.016 -- --
 1998 -0.030 -- --
 1999 -0.020 -- --
 2000 -0.058 -- --
 2001 -0.045 -- --
 2002 -0.011 -- --
 2003 -0.011 -- --
 Very Small -0.007 -- --
 Small -0.007 -- --
 Medium -0.005 -- --
 Large -0.006 -- --
 Very Large -0.007 -- --
 Public -0.006 -- --
 Private -0.007 -- --
 Foreign -0.007 -- --
 India -0.007 -- --
 Pakistan -0.007 -- --
 Sample Mean -0.007 -- --
 St. deviation -0.023 -- --

B. Marginal Variance
(Risk) Effects
 1985 -0.109 -1.406 0.400
 1986 -0.070 -1.170 0.488
 1987 -0.075 -1.722 0.436
 1988 -0.145 -3.631 0.619
 1989 -0.058 -0.442 0.642
 1990 -0.063 0.104 0.617
 1991 -0.017 0.031 0.377
 1992 -0.019 0.190 0.751
 1993 0.008 0.197 0.780
 1994 -0.012 0.952 0.752
 1995 -0.012 0.039 0.321
 1996 -0.008 1.526 0.794
 1997 -0.006 -0.019 0.678
 1998 -0.005 0.055 0.780
 1999 0.006 0.085 0.623
 2000 0.009 0.451 0.707
 2001 -0.006 0.011 0.748
 2002 -0.007 0.065 0.691
 2003 -0.011 0.082 0.650
 Very Small 0.000 0.007 0.741
 Small -0.001 -0.107 0.631
 Medium 0.002 -0.030 0.608
 Large 0.011 -0.446 0.593
 Very Large -0.196 1.329 0.564
 Public -0.092 -0.807 0.580
 Private 0.000 0.011 0.665
 Foreign 0.000 -0.053 0.739
 India -0.043 -0.384 0.633
 Pakistan -0.005 0.168 0.721
 Sample Mean -0.034 -0.346 0.653
 St. deviation 0.041 1.079 0.132

Note: TME is the total marginal effects computed as
summing up input and output marginal effects.

Table 4 Frequency Distribution of Technical Efficiency

Percentage
Efficiency
Interval Frequency Percentage

10-50 273 16.24
50-60 243 14.46
60-70 428 25.46
70-80 459 27.31
80-90 206 12.25
90-100 72 4.28

Table 5

Pearson Correlation Coefficients (Figures in Brackets
are Significance Levels)

 Characteristics

 Time Size TME

Time 1.000
 (0.002)
Size -0.075 1.000
 (0.002) (0.017)
THE 0.028 -0.058 1.000
 (0.255) (0.017) 0.000
Output 0.019 -0.057 0.991
 (0.444) (0.019) 0.000
Input 0.148 -0.129 0.142
 (0.000) (0.000) (0.000)
Efficiency 0.475 -0.447 0.023
 (0.000) (0.000) (0.351)

 Marginal Variance (Risk Effects)

 Output Input Efficiency

Time

Size

THE

Output 1.000
 (0.774)
Input 0.007 1.000
 0.774 (0.000)
Efficiency 0.017 0.146 1.000
 (0.499) (0.000) (0.000)