Previous Page Table of Contents Next Page


IX. Stochastic Profit Frontier Estimation for Broiler


9.1 Estimation of Environmental Equation

The estimation of the environmental equation has two objectives. The first one is to estimate effects of various factors that affect the farm spending on pollution abatement activities. The second objective is to provide an instrumental variable (IV) for the stochastic frontier estimation in section 9.3.

In this estimation, the dependent variable is the extra (marginal) costs that each farm spent specifically for pollution abatement activities. These costs include machine or equipment bought/used for this purpose (e.g., for incineration) and extra labor cost that the farm have to pay to hire additional workers specifically for these tasks.

9.2 Explanatory Variables

The explanatory variables include three group of variables, the operator's characteristics, location of farms, and farm characteristics. The operator's characteristics include the operator's gender, age, and years of experience in this farm, whether the farm had been run by the operator's parent(s) before s/he has undertaken it, the operator's (or spouse's) years of education (the greater number of the two). Farm location variables include distance to community, distance to waterway, and broiler density in the one kilometer radius estimated by the respondents. Other farm characteristics include farm size (defined by number of chicken raised per year), whether the farm has fishpond or other crops besides broiler, and whether the main operator has also engaged in any non-farm occupations.

Based on the cost measurement-which recorded marginal cost instead of full cost, about one-fourth of farmers reported having no extra environmental cost. Since these cases are likely to be censoring data, we use Tobit estimation instead of ordinary least square regression in estimating the pollution abatement equation.

Only samples with full information were used in the estimation (131 out of 170 cases). The estimation result is presented in Table 9.1 below. The estimation result, shown in Table 9.1, indicates that large farms tend to incur more pollution abatement cost per bird than smaller farms. Part of this could be resulting from the manner by which the costs are recorded (as discussed above). It is, however, also possible that the higher animal concentration in large farms would require more effort to deal with the pollution abatement, as the area might have limited absorptive capacity. The last possibility is that the large farms tend to be more environmental friendly as it is usually under more scrutiny by the neighbors and the local authorities and possibly face tougher regulation or internal control.

Farms with joint crop activities incur more cost compared with those who sold all the manure. This is partly because buyers often incur some cost of taking the manure-which might be internalized in the manure price. Farms that have the chicken house over the fishpond also incur practically no cost in dealing with the manure. However, farms that use manure on their own crops tend to incur more cost of gathering it and prepare it to be used on the farm.

As for other farmer's and farm characteristics, older farmers and farms in "broiler zone" tend to spend less in abatement costs. It is conceivable that lengthy established farms, especially in traditionally broiler areas, might be in a firmer position on how their rights to do business there. It is also possible that younger farmers are more sensitive to environmental issues.

9.3 Estimation of Stochastic Profit Frontier Function

9.3.1 Specification

Dependent variable: Natural logarithm of profit per chick

Explanatory variables: in three groups

The last variable is included to capture effects of farm's pollution abatement cost on her profit. Since, however, the variable is most likely to be an endogenous one, we needed an instrumental variable for that task. In this case, we regress the actual farm spending on a set of variables and use the predicted value of the regression as the instrumental variable (see Section II).

It should be noted that two important variables that are hypothesized to affect farm's profit per-kilogram were left out from the equation, namely, farm size and contractual arrangement, since they are conceivably endogenous variables that are often determined jointly with other choice variables or likely to correlate with the error terms of the profit frontier equation.

9.2.2 Assumptions Employed in Estimating Stochastic Profit Frontier Function

We made a number of assumptions and adjustments in our estimation. The most drastic one was to make price adjustments based on the following assumptions:

Both assumptions are consistent with what happened in the second half of 2002, where some batches of Thai chicken exported to the EU were found with traceable amount of banned antibiotics like Nitrofurans. As a result, the EU imposed a universal inspection scheme for every batch of chicken imported from Thailand. Some Thai exporters also withdrew some of the shipments that were in transit. Domestically, the chicken price dropped substantially, at least in a two-batch interval. This is consistent with what many farmers reported about price drop in the study period (late 2002). Chareon Pokaphand Food (CPF), which is one of the leading broiler exporter, has recently announced a drop of profit in the first quarter of 2003, citing a 30 drop of export price of broiler. At the same time, farmers reported feed and drug prices had increased significantly in the second half of 2002.

To compensate for the unusual period of study, we made these adjustments to every firm equally, percentage-wise[114]. These adjustments also help us solve problem of negative profits reported by a number of broiler farms, which would cause estimation problem as the dependent variable is logged.

Similar adjustment was made for medicine cost, where only last-batch cost was recorded. The farmer was asked, however, of whether the level of drug use during the last batch was considered normal. If the answer was no, the farmer would be asked for the percentage of deviation from normal use level. This figure would then be used to calculate the normal level of medicine use.

Other adjustments that were made for this estimation was to substitute missing values of some explanatory variables by the mean values of non-missing samples drawn from the same province. The numbers of cases that need this remedy were between 19 (for feed price) to 39 (out of 170 samples).

Finally, the samples consist of 29 wage-contract farmers where compensations were provided on per-bird (or per kilogram) basis. Although this type of contract does not have output price, the return each farm received appears to vary substantially. To make them comparable with output price faced by other farmers, we imputed their output prices by subtracting the average output price per kilogram received by the forward-contracted farmers in each province by the average compensation (per kilogram) that the wage-contract farmers in the same province received. The difference is then added to the actual rate of compensation that each wage-contract farmer in that province received. This would keep the price variation among the wage-contract farmers, and, at the same time, make their imputed compensation comparable to those of the independent and forward-contract farmers.

9.3.3 Estimation Result

The estimation was done by employing FRONTIER4.1 program written by Tim Coelli (as described in Coelli 1996). Although Coelli's Model 2 was written for production frontier, it could be applied for profit frontier as well. The result is showed in Table 9.2, along with the OLS estimation.

Based on the LR test, the stochastic frontier estimation is statistically different from the OLS estimation. The estimated value of Gamma g=sU2/(sV2+sU2) of 0.978 is also significantly different from zero, suggesting that the auxiliary equation (the technical efficiency equation) play an important role in the estimation of the frontier function.

The estimation indicates that the farm profit would increase with the output price. On the other hand, an increase in feed price would decrease profit per bird significantly. Day-old-chick price is not statistically significant. As for fixed factors, after scaling by farm size (using number of chicken in the last 3 batches), the estimation indicates that farms that have high ratios of fixed factors per chick are less profitable than those that have lower ratios. The feed conversion ratio (FCR), which represents technological difference in various farms has the expected negative sign, as less efficient farms (in terms of having higher value of FCR) tend to make less profit. All results are similar for both OLS and MLE estimations.

As for the technical efficiency equation (u, which in fact represent inefficiency, since the value of u would increase when the firm is far below the frontier. Based on our estimation, farms run by a female owner/operator tend to be more technically efficient. Farms that operated by younger operators are also more technically efficient. Location-wise, farms that are closer to community tend to be more technically efficient, which could be resulting from facing lower transaction cost--and probably get higher output price, than farms in more remote area. Also, farms in broiler areas (judging by animal density in one kilometer radius) tend to be more efficient. It is plausible that the farms are in traditional broiler area so that the farm owners/operators are experienced and knowledgeable. The environmental variable (predicted value of the environmental cost per bird incurred by the firm) was not statistically significant.

9.3.4 Farm Size and Technical Efficiency

Although farm size was not included in the estimation as an explanatory variable due to its endogeniety, it is interesting to see if technical efficiency is different across size. Table 9.3 below show mean technical efficiency across farm size. From Table 9.18, the small farm (with 5,000 or less broiler stock) tended to be the least efficient group. The very large farms (with more than 20,000 brolier stock) and the medium sized farms (5,000-10,000 broiler stock) tended to be most efficient, followed by the large farms (with 10,000-20,000 broiler stock).

9.4 Conclusion

The broiler industry in Thailand is now dominated by about a dozen of integrators. Integrators usually mix their own farming with contract farming-- which has become the most common mode of operation in the broiler business in the past two decades. The contracted farms are dominant in numbers, and in some periods, in number of broiler raised as well. The domination of the contracted farms is indicated by our samples. However, our samples might under-represent integrators' farms, which are small in number but are huge in sizes.

Among our samples, most of which are contracted farms, large farms tend to get better terms of contract, e.g., contract duration, output and input prices. The better terms are likely to come from economies of scale in transaction. In production, large farms are more inclined to rely on closed housings with evaporative cooling system, and get better performances, e.g., better feed conversion ratio, lower broiler mortality rate, and higher return per bird, even though the large farms spend more per bird to curb the environmental problems.

The estimation of the pollution abatement equation suggests that larger farm is more likely to incur more pollution abatement costs per bird. It is not clear whether these farms are subject to more stringent control by the local authority or contractor or not. It is also possible that the manner by which the environment-related questions were asked (i.e., asking for the extra costs that each farm spent specifically for pollution abatement activities) has played a part on this result.

Other farm characteristics, such as owner/operator age's and broiler density in one-kilometer radius have negative effects on the amount that the farm spent on pollution abatement activities per chick. It is conceivable that long-settled farms in typical broiler area are under less pressure from the community than in a new farm in new area.

As for the profit frontier function, the output and major input prices have the expected sign. As one would expect, feed price would affect farm profit significantly. The estimation also suggest more capital intensive farms-in terms of fixed cost per kilogram of chicken produced--are likely to make less profit per bird than the farms that could use raise more chicken with the same amount of fixed inputs. Location-wise, other things being equal, farms in concentrated broiler area tend to be more efficient, which could be resulting from fiercer competition among farms in such area. It might also be the case that those are traditional broiler areas where most farms have more experienced owner-operators. Farms that locate near a community tend to make more profit. However, when experience is controlled, farms run by older farmer-operations are likely to make less profit than younger farmers. Thus far, farm spending on pollution abatement activities have not yet affect the farm's efficiency significantly. This is supported by the fact that most broiler farms incurred very little cost in pollution abatement. In reality, broiler-raising does not create too many pollution problems either, which would be apparent when compared with other livestock production nowadays.

Since larger farms tend to get better deals on both output and input prices, as well as has lower annualized fixed cost per kilogram of chicken produced, they are more likely to be make more profit-both per bird and overall. Judging from these fact it is likely that when the broiler market is down-as in the second half of last year (2002), there would be little room for small farms, most of which stay afloat even when family labor cost was not imputed in the cost side. Only when external demand picks up-as in the second quarter of this year, the small farms would get a better price to make up for earlier loss.

However, for the farms that operate at a small scale-i.e., less than 10,000 birds at a time, it is not always the case the new technology (evaporative cooling house) would bring about more profit. For that scale, our data indicate that farms with evaporative houses have much higher fixed cost per bird than those with traditional housings, and could end up with less profit (or more loss). In this respect, as long as the integrators are willing to contract out small farms, these farms' best chance would be sticking with old low-technology low-investment strategy instead of investing in the new and more expensive one.

Table 9.1 Tobit Estimation of Environmental Equation

Dependent variable: expenditure on environment cost per cow (x.01 Baht/bird)

Farm Size (number of chicken per year)

.2833653781E-04

4.944*

Female Dummy

.2656479251

0.322

Ln Age

-1.990068019

-1.097

Maximum Yr of Education (operator or spouse)

-.1368307712E-02

-1.124

Years of experience

-.7173616269E-01

-0.833

Dummy if farm had been in family before the operator

.2862873502

0.122

Distance to village

.8326902883E-04

0.063

Distance to river

.8405355543E-03

0.897

Broiler density in one km radius

-.8290027587E-05

-1.654*

Dummy if farm has fish pond

.4912260006E-01

0.044

Dummy if also crop farm

2.231543275

2.211*

Dummy if operator has non farm fish pond

1.1057397

0.9646

Constant

5.359978977

0.805

Number of observations

131


LR Test chi2 (d.f.=12)

138.10*


*statistically significant at level 0.05 (one tail) or less

Table 9.2 Stochastic Profit Frontier Estimation: Broiler

a) OLS Estimation




Coeeficient

Asym-t stat

Constant

-2.1424

-1.824*

Ln Output Price

2.4958

5.594*

Ln Feed Price

-1.2468

-3.540*

Ln Day-Old-Chick Price

-0.0804

-0.424

Ln Capital Cost per chick

-0.1071

-3.310*

Feed Conversion Ratio

-0.4605

-2.970*

Log likelihood function

-63.49


b) MLE



Profit Frontier Function



Constant

0.63826

8.120*

Ln Output Price

1.59261

5.731*

Ln Feed Price

-1.17222

4.030*

Ln Day-Old-Chick Price

-0.02033

-0.156

Ln Capital Cost per chick

-0.05418

-2.509*

Feed Conversion Ratio

-0.26386

-2.656*

Technical Inefficiency (U)



Delta (Constant)

-9.23390

-2.215*

Female (dummy variable)

-0.56420

-2.085*

Ln Age

3.65030

2.714*

Ln Years of Education

0.23911

1.044

Ln Density of chicken (1 km radius)

-0.79321

-3.336*

Ln Distance to Community (km.)

0.27086

2.548*

Env cost incurred ('0.01Baht per chick) (predicted value)

-0.09780

-1.333

Sigma squared

0.83621

3.849*

Gamma (g)

0.978

203.41*

N


170

Log likelihood function


-3.3066

LR test (one-sided)


133.60*

Number of restrictions


8

*statistically significant at level 0.05 (one tail) or less.

Table 9.3 Mean Technical Efficiency by Farm Size

Farm Size

N

TE

Mean Adjusted Profit (Baht/Kg)

Mean unadjusted Profit (Baht/Kg)

1-5,000

74

.777

10.14

-0.01

5,001-10,000

51

.840

11.54

1.41

10,001-20,000

27

.811

11.83

1.63

20,000+

18

.851

13.36

3.21

Total

170

.809

11.17

1.02

Source: From survey and frontier estimation


[114] Since the dependent variable is logged, the effect of these adjustments was close - although not equivalent - to adding a positive value to the constant term of the regression.

Previous Page Top of Page Next Page