# 8.4 Explanation of Farm-Specific Profit Inefficiency in Terms of Differential Transaction and Environmental Mitigation Costs

The previous chapters and sections of this chapter now yield the data and methodology necessary to formally test hypotheses 4 and 5 from Chapter 1. For convenience, these are repeated here.

• Hypothesis 4: The relative profit efficiency of large-scale farms is more sensitive to environmental externalities than is the case for small farms.

• Hypothesis 5: Profits of small-scale producers are more sensitive to 'transaction costs' than are those of large-scale producers.

Understanding the determinants of relative profit efficiency clarifies which areas require policy intervention and improved technology and institutions in order to support the competitiveness of smallholder livestock farming. The measures of relative profit efficiency reported in the previous section are computed from the residuals of the stochastic profit frontier for the commodity and country concerned. An example of such a frontier is given for broiler farms in the Philippines in Table 8.5a. Any coefficient reported in this study is statistically different from zero at least a 90 percent confidence level or better.[32] To paraphrase the discussion in Chapter 4, the coefficients in Table 8.5a are MLE estimates of the influence of prices and farm fixed resources on maximum profit that a farm in the category concerned can make. For independent and price contract farms, for example (column two), a one percent rise in the price of broilers, others things equal, is associated with almost a 5 percent rise in the per unit profit; a one percent rise in the feed price is associated with a fall in unit profit of less than 5 percent (note that feed price shows up in more than one place in the frontier).

With minor exceptions here and elsewhere, the coefficients of the frontiers are of expected sign and magnitude. These coefficients are of only indirect interest to the present study, because our main reason for estimating the frontier is not the analysis of production relationships per se, but the estimation of what the ideal level of unit profit would be for each farm. Plausible coefficients suggest that the estimated frontier is usable for our purposes. The previous section used the results from the stochastic profit frontier analysis to compare relative profit efficiency across classes of farms. The present section will examine results from the second stage of the model in Chapter 4. This regresses the residuals from the first stage (measures of the amount that a given farm falls below the frontier ideal, or (Y* - Yi) in Figure 4.1, against a series of proxy variables for farm-specific transaction costs/policy distortions and our measure of farm-specific environmental mitigation (instrumented where appropriate).

Results for broilers in the Philippines are given in Table 8.5b. When the analysis is limited to independent and price contract farmers only (second column), the only variable to significantly explain differences in profit efficiency within the sub-sample is access to credit for feed. A negative coefficient means that the variable reduces inefficiency. In this case, access to feed credit explains why some farms are closer to the frontier. All the other variables could be important elements of farm efficiency, but there is too much or too little variation in their role across the sub-sample that it is not possible to estimate a coefficient with the require degree of statistical confidence.

Clear results for hypothesis testing come from the last two columns of table 8.5b. For large farmers, most of whom are independent, greater age, higher education, having had training, and access to feeds credit are all associated with greater relative profit efficiency (decreased inefficiency). For smallholders as a whole, greater education of the household head is associated with decreased profit efficiency. Rather than an indictment of the educational system, this is probably due to the fact that more educated sample small-scale operators are probably primarily engaged in other income ventures, with livestock being a sideline. The two major explainers of greater smallholder efficiency in the Philippines sample are having access to phone service and access to feed credit.

Interestingly, the impact of including the measure of environmental mitigation ("environmental capture" in Table 8.5b) in the explanation of relative profit inefficiency is precisely opposite for smallholders and large-scale farmers. Having a higher degree of environmental mitigation effort per unit of output is associated with being less profit efficient for smallholders, and more efficient for large farmers. The effect is not significant in the pooled regression involving both sub-samples. The interpretation of these results is that within the class of smallholders, those who pursue more environmentally sustainable practices in manure and dead animal disposal pay a price in terms of competitiveness, but the reverse is true for large farmers. Some smallholders can get away with throwing manure over the back wall in the river instead of spreading it, and make more money per unit as a consequence. This is less of an option for large farms, and in addition, most likely those who follow better practices have lower mortality rates and other efficiency gains relative to large farmers who are less careful.

Table 8.6, on Thai broilers, gives both the frontier results and the results explaining relative profit inefficiency in the same table. Results in the bottom half of the table suggest that the most efficient broiler farms are run by younger, female decision- makers, and are located in areas with a higher density of chickens, closer to population clusters. Differences in environmental mitigation behavior do not affect differences in relative profit efficiency within the sample. In the Indian broiler sample (Table 8.7), the availability of family labor, wage rates, and the price of broilers are the keys to smallholder profitability; larger farms need to worry primarily about labor and feed costs. The only clear driver of relative efficiency for smallholders is that greater effort for environmental mitigation lowers relative profitability, whereas it has no effect on the relative profitability of large farms. Experience contributes to greater efficiency on large farms, but not on small.

On Brazilian layer farms (Table 8.8), regional differences are paramount. The least profit efficient farms have operators with higher educational levels (probably a similar non-farm occupation explanation as in the Philippines), work in areas of higher concentrations of animals, closer to big cities, and make a relatively greater effort to mitigate environmental externalities. The least profit efficient Indian layer farms (Table 8.9) work in the South, are younger, and make the greatest relative effort at mitigation of environmental externalities. The picture is clearer, however, if the large-scale layer sample is considered in isolation. Among them, the most profit efficient are those who do the most per unit for environmental mitigation (similar to large-scale broiler producers in the Philippines), and are the best educated. In this case, the higher education is more likely to be used to support the layer enterprise as opposed to commuting to a non-farm job.

Table 8.5a Profit efficiency estimated and explained on Philippines broiler farms-stochastic profit frontier

 EXPLANATORY VARIABLES Pooled Independent and price contracts Contract fee Small-holders Large/Commercial Constant -4.535 21.896 4.790 -5.125 -31.514 Broiler price 5.551 4.849 - 4.57 8.616 DOC price ns ns - ns -3.139 Feed price ns -5.120 - 0.60 -3.167 Wage rate -3.527 -5.502 -0.620 -2.06 3.403 Lowest interest rate ns -6.265 ns 0.10 ns Buildings and eqpt/Q ns -4.700 ns ns ns Family labor/Q 1.741 2.057 0.334 1.075 -1.702 Land/Q ns -0.122 ns 0.012 ns FCR ns ns -0.593 -0.793 -0.397 Family labor x wage/Q -0.775 -0.924 -0.162 -0.457 0.705 Buildings and eqmt/Q x wage rate ns 1.658 ns 0.156 ns Buildings and eqmt/Q x inter rate ns 0.954 ns -0.34 ns Feed price x inter rate ns 4.232 - -0.18 ns Fee contract Dummy 23.007 ns - 17.64 17.618 Mean efficiency 0.578 0.506 0.839 0.595 0.827 log likelihood function -69.06 -55.079 18.060 -36.85 21.370

Note: "ns" means statistically insignificant at 10 percent.
Source: Costales, A., et. al., Annex I.

Table 8.5b Profit efficiency estimated and explained on Philippines broiler farms- explainers of inefficiency

 EXPLANATORY VARIABLES Pooled Independent and price contracts Contract fee Small-holders Large/Commercial Constant ns ns ns ns 2.445 Age ns ns ns ns -0.842 Educ ns ns ns 2.512 -2.058 Experience in poultry ns ns ns ns 0.660 Formal/informal training ns ns ns ns -1.453 Picked up services by buyer ns ns ns ns 0.245 Has phone services (Info) -1.535 ns ns -5.38 0.245 Environmental capture ns ns ns 0.234 -0.237 Feeds credit dummy -3.782 -3.132 - -6.39 -2.340

Note: "ns" means statistically insignificant at 10 percent.
Source: Costales, A., et. al., Annex I.

Table 8.6 Profit efficiency estimated and explained on Thai broiler farms

 Coefficients Stochastic Profit Frontier Constant 0.63826 Ln Output Price 1.59261 Ln Feed Price -1.17222 Ln Day-Old-Chick Price n.s. Ln Capital Cost per chick -0.05418 Feed Conversion Ratio -0.26386 Explainers of Inefficiencies Delta (Constant) -9.23390 Female (dummy variable) -0.56420 Ln Age 3.65030 Ln Years of Education n.s. Ln Density of chicken (1 km radius) -0.79321 Ln Distance to Community (km.) 0.27086 Env cost incurred ('0.01Baht per chick) (predicted value) n.s. Sigma squared 0.83621 Gamma (g) 0.978 N 170 Log likelihood function -3.3066 LR test (one-sided) 133.60 Number of restrictions 8

Note: "n.s." means statistically insignificant at 10 percent.
Source: Poapongsakorn, N., et. al., Annex IV.

Table 8.7 Profit efficiency estimated and explained on Indian broiler farms*

 Small Large Pooled Stochastic Profit Frontier Price of chicks -0.59 -1.30 -0.61 Wage Rate -2.72 -2.02 -1.64 Price of feeds -1.69 -1.42 -1.22 Price of output (Broiler) 3.49 n.s. 3.91 Family labour/Q 1.85 n.s. n.s. Value of capital stock/Q 0.03 n.s. 0.03 Slope of labour housing n.s. n.s. n.s. Wage Rate x Family Labour -0.49 -0.36 n.s. Feed Conversion Ratio -0.33 n.s. -0.30 Small vs. Large units (Dummy) -- -- 0.46 Constant 12.79 -4.85 1.49 Explainers of Inefficiencies Constant -21.42 -2.67 -28.98 Age of the decision maker n.s. 0.65 n.s. Information source (dummy variable = 1 for radio, TV, newspaper) 3.60 1.88 n.s. Region (Dummy, North = 1) -1.30 n.s 2.84 Pollution abatement costs n.s. 0.32 n.s. Education of Decision Maker 6.32 n.s 5.36 Has access to credit, dummy n.s. n.s n.s. Output Market Distance - n.s. n.s. n.s. Experience 2.47 -0.38 0.69 N 93 42 135

*Contract farmers not included
Note: "n.s." means statistically insignificant at 10 percent.
Source: Mehta, R., et. al., Annex II.

Table 8.8 Profit efficiency estimated and explained on Brazilian layer farms

 Coefficients Stochastic Profit Frontier Constant -101.202 Feed Conversion -1.443 Feed Price -1.257 Price of Hired Labor -0.169 Price of Electricity n.s. Price Freight n.s. Agricultural Land 39.232 Family Labor n.s. Capital -0.060 Explainers of Inefficiencies Constant -16.126 Length of Time Decision Maker in this Farm n.s. Level of Education for Decision Maker 3.253 Age of the Decision Maker n.s. Length of Time Decision Maker in Activity n.s. Animal Concentration in Regions 1.843 Animal Concentration on the Farm n.s. Environmental Cost 1.345 Distance of the City 1.136 Information Index n.s. State Dummy SP -10.336 State Dummy PR -4.527 Sigma-squared 2.743 Gamma 0.998

Note: "n.s." means statistically insignificant at 10 percent.
Source: Camargo Barros, G.S., et. al., Annex V.

Table 8.9 Profit efficiency estimated and explained on Indian layer farms

 Small Large Pooled Stochastic Profit Frontier Wage Rate 7.30 n.s. n.s. Price of feeds -1.69 -1.53 -1.46 Price of Egg 2.04 1.00 0.93 Family labour/Q -6.07 n.s. n.s. Value of capital stock/Q n.s. n.s. n.s. Slope of labour housing n.s. 0.02 n.s. Wage Rate x Family Labour 1.62 n.s. n.s. Wage Rate * SCALE DUMMY n.s. -- 0.60 Feed Conversion Ratio n.s. n.s. n.s. Constant -46.53 0.69 10.64 Explainers of Inefficiencies Constant 0.12 3.56 4.64 Age of the decision maker n.s. -3.43 -3.36 Region (Dummy, North = 1) n.s. n.s. -6.80 Pollution abatement costs n.s. -6.38 6.29 Education of Decision Maker n.s. -3.17 n.s. Has access to credit, dummy n.s. n.s. n.s. Information source (dummy variable = 1 for radio, TV, newspaper) n.s. 0.81 n.s. Output market distance n.s. 0.88 n.s. N 62 96 158

Note: "n.s." means statistically insignificant at 10 percent.
Source: Mehta, R., et. al., Annex II.

Swine production in the Philippines is analyzed in Tables 8.10a and 8.10b. Several patterns emerge from the analysis of the determinants of relative profit inefficiency in Table 8.10b. First, the transaction cost variables proxied by the independent variables show-which are correlated with household access to information and assets-matter more to smallholders than to large-scale swine farmers. Second, access to investment credit is a distinguishing feature of relative efficiency in all cases, but feeds credit access seems to contribute to relative inefficiency, a surprising and not very plausible result that needs further investigation. Third, having access to information is an important driver of relative efficiency for all farm classes except fee contractors, who presumably are not making the key production decisions in any event. Fourth, farmers who sell to itinerant traders (viajeros) are typically less efficient that farmers who have other marketing arrangements. Fifth, fee contractors are more efficient than independent and price contract farmers, other things equal, confirming the results from the previous section. Sixth, in all cases, greater efforts to promote environmental mitigation are associated with increased profit efficiency.

In the Brazilian sample (Table 8.11), the most efficient swine farms were headed by more educated decision-makers that had longer experience in swine, and were located far from big cities, in regions with high animal concentrations and high animal density on the farm. Engaging in environmental mitigation contributed strongly to reducing relative profit efficiency.

The results for dairy farms are shown in Tables 8.12 and 8.13. Disaggregated results for small and larger farms in India show considerable homogeneity in the driving forces of relative profit efficiency for small and medium dairy farms. This is not surprising, since having more than 10 milk cows counts as a large dairy farm in the Indian sample. A striking result is that the hypothesized explainers of relative profit efficiency have quite strong effects for small and medium scale farms, but no effect on relative profit efficiency within the large and commercial farm sample.

Table 8.10a Profit efficiency estimated and explained on Philippines swine farms-stochastic profit frontier

 EXPLANATORY VARIABLES Pooled (N=207) All Independents (N=154) All Contracts (N=53) All Smallholders (N=110) All Large-scale (N=97) Constant 3.564 ns 2.412 ns ns Price of hogs 0.685 1.233 1.08 0.789 1.506 Price of piglets 0.727 0.738 - 1.110 -0.495 Price of weanlings -0.119 -0.119 -0.266 -0.116 - Price of feed -0.687 -0.718 -0.814 -0.696 -0.544 Contract dummy -0.405 - - -0.373 3.227 Wage rate ns ns ns ns ns Free board and lodging dummy ns ns ns -0.106 ns Lowest interest rate ns ns ns ns ns Value of inventory/Q ns ns ns 0.018 ns Buildings and eqpt/Q ns ns ns ns ns Family labor/Q ns ns ns -0.379 ns Farmsize (in ha)/Q ns ns ns ns ns FCR ns ns ns ns -0.115 mortality rate ns ns -0.0178 ns ns Family labor x wage rate ns ns ns ns ns Buildings and eqpt/Q x lowest I.r. ns ns ns ns ns farrow-wean dummy ns 2.315 - ns 8.695 combined farrow-wean and farrow-finish -2.986 -3.012 - -4.599 2.555 Very large dummy - - - - ns

Note: "n.s." means statistically insignificant at 10 percent.

Source: Costales, A., et. al., Annex I.

Table 8.10b Profit efficiency estimated and explained on Philippines swine farms-explainers of inefficiency

 EXPLANATORY VARIABLES Pooled (N=207) All Independents (N=154) All Contracts (N=53) All Small-holders (N=110) All Large-scale (N=97) Constant -6.594 -1.978 ns ns 2.197 Age ns ns ns ns ns Experience in hog business ns -0.426 0.102 ns ns Education ns ns ns ns ns Formal/informal training 2.052 2.833 ns 1.799 ns Feeds credit dummy ns 1.086 0.271 2.220 ns Capital credit dummy -1.448 -0.986 -0.161 -1.945 ns Information -4.972 -3.078 0.702 -3.665 ns Buyer is viajero 4.12 2.08 - 2.857 ns Environmental cost ns -0.77 -0.106 -1.217 ns Fee contract dummy -1.391 - -0.747 - -1.838 Very large dummy - - - - Region dummy (Bukidnon=1) ns -1.41 ns ns -0.703 sigma squared 2.085 1.357 0.0127 1.583 0.203 gamma 0.994 0.992 0.7 0.997 0.984 Mean efficiency 0.802 0.781 0.746 0.802 0.812

Note: "n.s." means statistically insignificant at 10 percent.
Source: Costales, A., et. al., Annex I.

Table 8.11 Profit efficiency estimated and explained on Brazilian swine farms

 Coefficients Stochastic Profit Frontier Constant 0.000 Feed Conversion -0.816 Price of Feed -0.677 Price of Hired Labor n.s. Price of Electricity n.s. Price of Environmental n.s. Price of Output 0.534 Dummy Complete Cycle (1=yes or 0=no) -0.218 Dummy Independent (1=yes or 0=no) n.s. Dummy Integrate (1=yes or 0=no) 0.102 Agriculture Land 43.883 Family Labor n.s. Capital n.s. Explainers of Inefficiencies Constant n.s. Length of Time Decision Maker in this Farm 0.203 Level of Education for Decision Maker -0.567 Age of the Decision Maker n.s. Length of Time Decision Maker in Activity -0.256 Animal Concentration in Region -0.388 Animal Concentration on the Farm -6.649 Environmental Cost 8.098 Distance of the City -0.325 Farm Distance to Nearest Neighbor 0.206 Information Index 0.643 State Dummy SC -1.205 State Dummy PR n.s. State Dummy MS -2.816 State Dummy MT -4.911 State Dummy GO -2.601 State Dummy RS n.s. Percent Share of Swine Production Total Income n.s. Sigma-squared 0.478 Gamma 0.986

Notes: "n.s." means statistically insignificant at 10 percent.
Source: Camargo Barros, G.S., et. al., Annex V.

This not to say that access to information and assets is not important for large farms; it only means that farm-specific differences in this area did not distinguish farms from each other with respect to profit efficiency. On the smallest farms, access to information and credit were the key drivers of relative efficiency. On farms with 4 to 10 milk cows, the key drivers were differences in education and information. Thai dairy farms (Table 8.13) exhibit the same pattern as the larger scale Indian dairy farms. The relative profit efficiency of Brazilian dairy farms (Table 8.14) is driven primarily by location, the degree of family involvement in the enterprise (good for efficiency), establishment of that farm in dairy, and access to information.

In sum, we can say the following with regard to hypothesis 4. Monetizing a measure of environmental mitigation, or internalization of negative externalities, helps explain why some farms are more or less profit efficient in the majority of cases. However the issues are different within the separate categories of large and small farms. Furthermore, differences in environmental mitigation do not seem to be strong explainers of differences in profit efficiency across sizes of operation. The brunt of evidence is that within large-scale operations for swine and poultry, greater effort for mitigation of environmental externalities seems to be associated with greater relative profit efficiency. This is clearest in the case of broilers and swine in the Philippines and layers in India. Interestingly, these tend to be mostly independent operations that are transitioning towards more industrial production, at least as compared with other samples studied. Results for smallholders are more mixed, particularly where contract and independent sub-samples are mixed, as in the Philippines. Most fee (or wage) contractors have to follow a standard set of environmental practices as part of their contract, and they resemble larger scale farms in this respect more than other smallholders.

Table 8.12 Profit efficiency estimated and explained on Indian dairy farms

 Small Farms Medium Farms Large and Commercial Farms Stochastic Profit Frontier Intercept -3.7117 n.s. n.s. Price of milk 1.8266 1.6231 0.9672 Price of fodder -0.5946 -0.4034 -0.5244 Price of feed -0.4091 -0.2726 -0.3270 Yield 0.8738 0.2859 0.3557 Family labor n.s. 0.2781 n.s. Wage rate n.s. -0.6128 -0.3903 Family labor x wage rate n.s. n.s. n.s. Land n.s. n.s. 0.0450 Building & equipments n.s. n.s. 0.1333 Land x Building & equipments n.s. n.s. 0.0359 Explainers of Inefficiency Intercept -3.9495 n.s. n.s. Age n.s. n.s. n.s. Education n.s. -0.1291 n.s. Distance from market n.s. n.s. n.s. Access to information -5.0416 -0.6850 n.s. Access to credit -1.0867 n.s. n.s. Environmental cost n.s. 0.7810 n.s. Zone dummy 3.5495 0.4525 n.s. Gamma 0.9314 0.4524 n.s. Log likelihood ratio -102.5239 23.9918 23.7967

Note: "n.s." means statistically insignificant at 10 percent.
Source: Sharma, V.P., et. al., Annex III.

Table 8.13 Profit efficiency estimated and explained on Thai dairy farms

 Coefficients Stochastic Profit Frontier Constant -0.5026865 Ln Output Price 1.535576 Ln Price of Concentrate Feed -0.5806623 Ln Price of Roughage n.s. Ln Capital Cost per kg milk n.s. Ln Farm Land per kg milk 0.1237507 Yield (monthly output of milk per number of milking and pregnant cows) 0.0022744 Explainers of Inefficiency Delta (Constant) n.s. Male operator (dummy variable) n.s. Ln Age n.s. Ln Years of Education (maximum year of operator's or spouse's education) n.s. Ln Distance to Community (km.) n.s. Ln Distance to Waterway (km.) n.s. Env cost incurred (Baht per cow) (predicted value) n.s. Sigma squared n.s. Gamma 0.953688 N 89 Log likelihood function -81.404 Wald Chi2 Test 33.97 Number of restrictions 6

Note: "n.s." means statistically insignificant at 10 percent.
Source: Poapongsakorn, N., et. al., Annex IV.

Table 8.14: Profit efficiency estimated and explained on Brazilian dairy farms

 Coefficients Stochastic Profit Frontier Constant -1.682 Milk Production by Cow in Lactation per Day 0.268 Humid Feed Price -0.071 Dry Feed Price -0.194 Medicine Price -0.063 Genetic Price 0.025 Electricity Price n.s. Hired Labor Price -0.048 Output Price 0.846 Membership of a Cooperative n.s. Agricultural Land n.s. Family Labor -0.122 Capital n.s. Value of Herd -3.541 Explainers of Inefficiencies Constant -3.691 Years of Dairy on That Farm -4.120 Experience of Owner in Dairy 1.311 Decision-maker is Owner -1.925 Decision-maker is Family Member -3.397 Decision-maker Experience in Dairy (years) n.s. Decision-maker lives on Property -1.703 Age of Decision-maker 4.614 Education of Decision-maker 2.005 Had Training? 1.369 Distance between the Farm and the City (km) n.s. Information Index -0.502 Environmental Cost n.s. State Dummy RS n.s. State Dummy SC 1.904 State Dummy PR n.s. State Dummy SP -2.748 State Dummy MG n.s. Sigma-squared 1.558 Gamma 0.998

Notes: "n.s." means statistically insignificant at 10 percent.
Source: Camargo Barros, G.S., et. al., Annex V.

The environmental mitigation variable did not seem to have much influence on relative profit efficiency in the Thai sample. Egg and swine producers in Brazil, smallholder swine producers in the Philippines, and large-scale broiler farmers in India that spent relatively more on environmental mitigation tended to have lower relative profit efficiency at the end of the day, other things equal. It is interesting to speculate whether these sub-samples operated in conditions where it was relatively easier to ignore environmental issues, or perhaps harder to follow environmentally sound practices because of land scarcity.

With regard to hypothesis 5, farm-specific transactions costs seem to matter greatly to explaining relative profit efficiency across farms in most of the sub-samples studied. This means that relatively greater difficulties in securing access to assets and information for smallholders is a prime explainer of differences in relative profit efficiency within their group, and between them and large-scale farmers.

The most notable exceptions-where farm-specific differences in transactions cost proxy variables did little to explain differences across farms in relative profit efficiency-occurred for dairy farms. It is likely that transaction costs for dairy almost all occur in the marketing chain and not at the level of production, at least in the Indian and Thai contexts. Feed is mostly forage (avoiding the high credit and quality-related transactions costs packed into using concentrate feeds) and the timing of sales is a foregone conclusion, viz. daily. This is quite unlike farmers of monogastrics, where the timing of sale is more discretionary (requiring information), much less frequent, but critical to profit margins. For those cases where transaction cost variables matter most to smallholder producers, the main issues appear to be access to telephone service and the market information that goes along with this, and access to credit for feed purchase.

 [32] More formally, the null hypothesis that the coefficient is zero is rejected with less than 5 percent of a chance of making error, using a two-tailed test.