Previous Page Table of Contents Next Page

X. Stochastic Profit Frontier Estimation for Layer

10.1 Estimation on Environmental Equation

The dependent variable is additional costs that each farm spends specifically for pollution abatement activities. The costs include machine or equipment invested for this purpose and shadow labor costs that farms pay for this purpose. It is standardized into 'per bird' variable, since it is the layers, not the eggs, which generate environmental problems in the farms.

The explanatory variables are divided into three groups, which are the owner/operator's characteristics, location of farms, and farm characteristics. The owner/operator's characteristics include his/her age, years of education, the social status of the owner/operator, and year of experience in the business. The experience in the business is measured by the age of the farm, since experience can be transferred from the owner/operator's parents. Farm location variables include distance to community, distance to waterway, and provincial dummies. The farm characteristic variables cover number of mature layer in the farm (representing the farm size), the flies management dummy (1 if manage, 0 if not), and the owner/operator's other sources of income, which are divided into incomes from other activities in agricultural area and incomes from activities out of agricultural area.

Tobit model is implemented instead of OLS in order to deal with missing data. The estimation result is in Table 10.1 below.

The regression result indicates that the sizes of the farms have no significant relationship with the environmental abatement costs farms pay. In one way, better environment could generates better income and reduce conflicts with their neighbors and local authorities; hence, farms in all sizes invest in the environmental abatement costs up to a level. In the other way, investing too much for environmental issues ensue extra costs; therefore, all farms avoid it.

For the farm location, peculiarly the farms locate close to communities will pay less to deal with environment. The result goes against the provincial dummies, which tell us that farms locate in a big city - Nakorn-Ratchasrima - tend to pay more to manage the environmental issues. Farms close to the waterways also tend to pay more to manage their environment. These latter two results indicate that local authorities and neighborhood still have influences over environment abatement costs of the farms.

Owners/operators' age, education and experiences have no significant effects to environment costs. Only their social statuses matter. The social statuses tend to reduce costs of environmental management. It may provide alternatives to shift environmental costs to the external economy.

10.2 Descriptive Analysis

Figure 10.1 - 10.4 show some interesting differences among farms across scale categories. They can also provide some basic ideas for developing the profit frontier in the next section.

Figure 10.1 below shows average egg price across production scales. It says that larger-scale farms receive less for each unit of egg. It does not tell us that larger farms produce less quality products, but, instead, smaller farms tend to involve more in marketing activities, e.g. sales, especially the small-scale farms. Larger farms, in turn, tend to implement more market intermediaries' services.

Figure 10.2 below shows average feed price per kilogram across farm scales. Feed cost is the most important item composing variable cost for a layer farm; hence, the lower price of the feed will provide a significant benefit to the farm. Larger farms tend to benefit from lower feed prices, since they tend to have more bargaining power to the suppliers of the finish feed and of the raw materials to produce mixing feed.

The following figure 10.3 reports the average annual eggs per bird across farm scales. This variable can in part represent efficiency in the operation of the farm, since it reflects the return that a mature layer will provide to the farm. However, this variable does not capture the information on the cost farm invests in a mature layer. Larger farms tend to operate with higher efficiency (in this meaning) than smaller farms.

Eventually, the information from Figure 10.1-10.3 found the resulting farms' profits shown in Figure 10.4 below. The profits are standardized into 'per 1000 eggs' term. The profit per 1000 eggs is at their peak at the medium-scale farms. Large-scale farms tend to profit least for each 1000 eggs. This does not mean that large-scale farms operate at least efficiency, instead as we can see from Figures 10.2 and 10.3 that they operate very efficiently comparing with farms in other scale categories. The lower profits should be explained by the lower prices received, which, as stated earlier, are caused by different marketing activities that farms in each scale category involve.

10.3 Estimation of Stochastic Profit Frontier Function

10.3.1. Specification

a) Dependent variable: Natural logarithm of profit per 1,000 eggs

The profit in this regression means to 'total revenue - total variable cost'. The profit is standardized to be per '1,000 eggs' because it makes the natural logarithm values to be positive for all farms.

b) Explanatory variables: Three groups

1. Output and input prices (in natural logarithmic form)

- Egg price (bath per 1,000 eggs)

- Feed price (bath per kg)

Price of 18-week-old layer (bath per bird). The farms can either buy the 18-week-old layers from their suppliers or raise one-day-old chick to be the 18-week-old. For the former case, the price instantaneously exists. For the latter, the price proxy is calculated from summing the price of one-day-old chick with feed and drug costs related with raising the chick.

- Wage rate. The question of labor wage rate does not appear in the layer questionnaire, while other proxies available are unreliable. Hence, the variable is dropped from the model.

- Feed conversion ratio (kg per 1,000 eggs). Since the output of layer farms are eggs, the feed conversion ratio is calculated by dividing the amount of feed farms used for mature layers to produce 1,000 eggs.

2. Fix factors (in natural logarithmic form standardized to be per 1,000 eggs)

- Capital (bath per 1000 eggs×year). It is the annual amortized cost per 1,000 eggs. The amortization assumed simple linear depreciation.

- Family labor. The variable is dropped from the model with the reason of possible measurement error and also possible bias generated from dropping wage variable from the model.

3. Technical efficiency auxiliary equation

- Female dummy (1 for female owners/operators, 0 otherwise).

- Owner/operator's age (years). The variable is in natural logarithmic form.

- Owner/operator's years of education (years). The variable is in natural logarithmic form.

- Housing area per bird (square meters per bird).

- Distance from farm to the nearest community (km.). The distance is estimated by the respondent. It is in natural logarithmic form.

- Environmental cost (Baht). Predicted value.

10.3.2 Corrections of Data

The most prominent problem occurred in the survey data is that the profit calculated in the stated meaning become negative. In order to fix the problem, some corrections are made on the belief that at least the farms should operate above their shutdown points.

For layer farms, the problem tends to be caused from two major factors. The first one is on the number of eggs sold in one day. The stocks of mature layers in many farms fluctuate along the year and the number of eggs sold varies with these stocks. With the question that tries to capture the yearly average eggs sold in one day, many farms respond too high or too low the number of eggs sold, comparing to the stocks of mature layers of the farms in the present time. Hence, for these 26 cases, the data on eggs sold in one day are substituted by the calculations that make use of the information on the 'annual number of eggs per one bird' and the 'present stock of mature layers' of the farms. These calculated figures provide more suitable and plausible results.

The second factor is on the feed cost. Many farms tend to report very high feed costs, which make their operating result negative. These negative figures tend to go against our stated belief and the information that they are still paying positive income tax. Since the information on feed price is typically available, our doubt goes to the amount of feed layers eat in one day that the farms report. Seventeen farms' data are substituted by the calculated figures that come from the information on 'the amount of feed the farms buy/mix in one month' and the 'present stock of layers' in the farms.

The other minor corrections are on the too high monthly electricity fees, wage expenditures, and drug fees. Standardized unit costs are informally surveyed and multiplied with the number of units used in the farms. These calculated figures would replace the survey data.

10.3.3 Estimation Result

The estimation is done through the program FRONTIER 4.1 written by Tim Coelli. The function is applied with Coelli's (1996) Model 2 for the case of production frontier. The result is shown in Table 10.2 below.

The LR test shows that the MLE stochastic frontier estimation is statistically different from the OLS estimation. The estimated value of Gamma of 0.9991 is different from zero, indicating that the auxiliary equation plays important role in the estimation of the frontier. Hence, concentration should be on the MLE estimates.

Output price is a major factor influencing the profit of the layer farms as it gives strong significant positive effect to the function, while feed price and the effective uses of feed are the main factors decreasing the farms' profits. Increase in feed price generates strong significant negative effect to the profit function and increase in feed per 1,000 eggs yields analogous effect. The prices of 18-week-old layer and capital costs possess insignificant effects. This may be the result of the long duration of the factor.

For the technical inefficiency equation, gender and age of the owners/operators of the farms insignificantly relate to the efficiencies of farms. Peculiarly, more years in school of the owners/operators significantly increases inefficiencies of farms.

Larger housing area for mature layers and higher expenditure in environment abatement costs can improve efficiencies to farms. This findings may go against original believes; however, they are empirical.

At last, if we look at the overall efficiencies of farms, we will see a big picture. The average estimated efficiencies across farm scales are shown in figure 10.5 below. The figure says that totally the small-scale farms operate least efficiently, while the large-scale farms operate most efficiently. Hence, for the current picture, the small-scale farms may need helps to make them still competitive in the industry.

10.4 Conclusion

The environmental equation from tobit estimation in this study indicates that the size of the farms do not affect significantly on the costs farms invest to abate the environmental problem. Instead, the farm location and the owners/operators' social status variables play more significant roles in determining environment abatement costs of the farms. In the other words, external factors play more important roles in determining these costs.

It comes to the question that whether the internal decisions in the farms affected by these environmental management costs. Hence, the costs are brought into profit frontier functions of the farms, in the part of auxiliary efficiency equation. To prevent spurious regression problem, the estimated values of the costs are brought to replace the actual costs as an instrumental variable.

The profit frontier estimation result shows that increase in environmental abatement costs can significantly improve efficiencies in farms' operation. The result may go against prior believes and this can lead to the policies of win-win between the owners/operators of the farms and the society.

In concerning on the efficiencies across farm scales, Figure 10.1 shows that large-scale farms can do better than medium-scale and small-scale farms. This finding can be partly strengthened by the average yields across farm scales shown in Figure 10.3 However, the improvements are quite marginal.

In addition, if we look at the average profit across farm scales (Figure 10.4), we will observe that the profits per standardized unit of the medium-scale and small-scale farms are higher than one of the large-scale farms. This can be explained by the facts that small-scale and medium-scale farms benefit from higher price by involving more in marketing activities than the large-scale farms. These benefits can offset the inefficiencies occurred and explain the existence of small-scale and medium-scale layer farms in the present time.

Table 10.1 Estimation of Environmental Equation






Ln number of mature layers



Ln owner/operator's age



Owner/operator's years of education (years)



Social status dummy



Experience of the farm (years)



Distance to the nearest waterway (km.)



Distance to the nearest community (km.)


* 1.58

Flies management dummy


** 3.43

Income from other sources in agricultural area (baht/month)



Income from other sources out of agricultural area (baht/month)



Provincial dummies

- Chon-Buri



- Chacherngsao



- Lop-Buri



- Nakorn-Ratchasrima


** 2.60



Uncensored observations


Log likelihood function


* statistically significant at level of 0.10 (one-tailed).
** statistically significant at level of 0.05 (one-tailed) or less.

Table 10.2 Regression Result on Stochastic Profit Frontier Function

OLS Estimation





** 1.813

Ln output price


** 7.113

Ln feed price



Ln 18-week-old layer price



Ln capital cost per 1,000 eggs



Ln feed conversion ratio



Log likelihood function


MLE Estimation




Ln output price



Ln feed price



Ln 18-week-old layer price



Ln capital cost per 1,000 eggs



Ln feed conversion ratio



Technical inefficiency (u)




Female dummy



Ln owner/operator's age



Ln owner/operator's years of education


** 2.884

Housing area per bird



Ln distance from farm to nearest community



Expected value of environment cost



Sigma squared


** 3.171






Log likelihood function


LR test (one-sided)


Number of restrictions


* statistically significant at level of 0.10 (one-tailed).
** statistically significant at level of 0.05 (one-tailed) or less.

Figure 10.1 Average Egg Price

Figure 10.2 Average Feed Price per Kilogram

Figure 10.3 Average Annual Eggs per Bird

Figure 10.4 Average Profit per 1,000 Eggs

Figure10.5 Average Estimated Inefficiencies Across Farm Scales.

Previous Page Top of Page Next Page