VII. Profitability and Efficiency of Indian Dairy Farms

The central problem in the neoclassical theory of production is efficiency in the allocation of resources. A producer is said to be efficient in resource allocation if the optimality conditions are satisfied. Similarly, a producer is technically efficient if maximum possible output is obtained from a given quantity of input. It is widely believed that there is substantial variation in economic efficiency and profitability across different farm sizes. These differences in relative performance can be due to differential transaction costs stemming from asymmetric access to assets and information between large and small farm sizes, differences in spending on environmental practices, or differences in access to policy subsidies. We hypothesize that these issues can be better explained in terms of relative economic efficiency, farm sizes, and economies of scale. If efficiency varies across farms, those with relatively more efficiency will be more profitable. If small farms have lower profits and are less efficient than medium and large farms, and government policy supports the large farms, the small farms will be driven out of the market. The reverse, however, does not mean that smallholders will not be displaced, because large farms can compensate for lower per unit returns with large production volumes.

7.1 Profitability of Dairy Farms

This chapter examines the profitability of Indian dairy farms in relation to farm size and relative economic efficiency. Specifically, it focuses on the impact of technical and allocative efficiency on the level of profit of farms of different sizes. First we use a profit function that corresponds to variable returns to scale. Second, technical efficiency is assumed to have deterministic and stochastic components. The exogenous variables explaining a deterministic component are viewed as determinants of technical inefficiency. We develop a single-step maximum likelihood procedure to obtain consistent parameter estimates and identify determinants of technical inefficiency. The two-step procedure has been defended by many authors (Kalirajan, 1991; Kalirajan and Shand, 1985; Ray, 1988), while others (Battese and Coelli, 1995; Kumbhakar et al., 1991,) have challenged it by arguing that the two-step procedure may give inconsistent parameter estimates and that farm-specific factors should be incorporated directly into the estimation of the production frontier, because such factors might have a direct impact on inefficiency.

To measure farm-level economic efficiency, we have applied the stochastic profit frontier methodology to primary data from a survey of 520 households in Punjab, Haryana, and Gujarat on the basis of a multistage stratified random-sampling technique.

Each observation included milk production per farm measured in liters per day, price of feeds and fodder, and liquid milk price. Labor used per farm included family labor measured in man-hour equivalents. The price of labor in terms of man-hour equivalents was computed on the basis of the wage rate paid by sample dairy farms for hired labor and the prevalent market wage rate for family labor. In measuring capital stock, we used the value of cattle sheds and other farm equipment used in various operations of dairy farming. To capture the effects of technological change, yield per animal was included as one of the explanatory variables in the model. In addition to these inputs, we used farm size and regional dummies as control variables. Farm size is defined by the number of dairy animals. Farm size is considered small if the number of dairy animals does not exceed 3, medium if the number is between 4 and 10, and large if the number exceeds 10. To control for regional effects, we divided the sample into two regions. Region 1 is the western state of Gujarat, and region 2 is the northern states of Punjab and Haryana.

We considered level of education of the dairy farmer as the determinant of technical inefficiency. There is debate about whether one should consider education as an input that enhances the allocative efficiency of a producer (Huffman, 1977; Schultz, 1975) or as human capital that enhances technical efficiency (Yotopoulos and Nugent, 1976). In this study, we used education (years of formal schooling) as a determinant of inefficiency. Age of the farmer was used as one of the variables that influences the inefficiency. Generally young farmers are believed to be more efficient than old farmers. To capture the impact of transaction costs, we used dummy variables for access to information and credit. For different levels of negative environmental externalities, expenditures on pollution abatement were used. Distance to market (in km) was used capture the impact of farmers' access to output markets and infrastructure. The maximum likelihood (ML) estimates of the parameters of the stochastic profit frontier were obtained by using the program Frontier 4.1 (Coelli, 1994).

Hypothesis 1: Small-scale producers have lower profits per unit of output than do large producers.

Before discussing the profit frontier results, we will address the question of relative profitability of different farm sizes and test the first hypothesis, that "small-scale producers have lower profits per unit of output than do large producers." Profits here do not take into account uncosted inputs such as family labor and environmental externalities. Traditionally, family labor at reservation wages much below market wages has been a source of competitiveness for smallholders. To test this hypothesis, we computed average profit (gross revenue minus total variable costs) per liter of milk, excluding family labor and environmental externalities, and compared across different levels of output. The results are presented in Figure 7.1.

The null hypothesis that small farmers have lower profits per unit of output is strongly rejected, as average profit per liter of milk showed an inverse relationship with farm size. The mean profit per liter of milk excluding family labor was highest (Rs. 2.45 per liter) on farms producing up to 10 liters of milk per day and lowest (Rs. 2.50 per liter) on farms producing more than 150 liters milk per day. However, there were regional variations in profitability across different farm size groups (Figure 7.2). Small farmers in the western region have higher profits than their counterparts in the northern region. In contrast, large farmers in the northern region have high profitability compared to those in the western region. One of the explanations for this differential pattern might be the strong presence in the western region of cooperatives, which provide an assured market for milk output irrespective of level of production and other production inputs and services. In the northern region, farmers are at the mercy of middlemen/milk vendors/dudhias, and small farmers have less bargaining power and might get lower milk prices as well as poor access to other inputs and services.

Figure 7.1 Mean profit (excluding family labor) per liter of milk across size groups (daily milk production in liters) in India

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Figure 7.2 Mean profit (excluding family labor) per liter of milk across different farm sizes (daily milk production in liters), by region

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

One may question why family labor should not be counted in calculating profitability because in some regions opportunity cost of labor is high. To address this question, we computed mean profits per unit of output by including family labor and compared across farms and regions. The results are presented in Figure 7.3. It is evident from Figure 7.3 that the relationship between farm size and profitability remained unchanged but the level of profits fell substantially for small farms, as small farms use mostly family labor for milk production activities. The mean profit per liter of milk was highest (0.50 rupees per liter) on farms producing less than 10 and 20-40 liters of milk per day and declined with the increase in production, with the exception of the 10-20 liter size class, where the profitability was lower compared to first and third size farms. The average profit per liter of milk declined from Rupees 1.37 without family labor to 0.43 rupees per liter when family labor was counted in the costs. This decline was more pronounced on small and medium farms, as they use mostly family labor for dairy farming activities.

Figure 7.3 Mean profit (including family labor) per liter of milk across different farm size groups (daily milk production in liters) in India

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Average profit per liter of milk was higher in the northern region than in the western region. However, average profit among small farms was higher in the western region, while among large commercial farms it was higher in the northern region. The average profit per liter of milk was. 0.66 rupees in the western region and. 0.44 rupees per liter in the northern region. The reduction in profitability by including family labor was higher in the northern region because the wage rates in the northern region were higher than in the western region.

These results clearly show that small-scale producers have higher profits than large producers both with and without taking into account family labor. Small-scale producers in the western region have higher profitability, while large farms in the northern region have higher profitability. The average profitability (without family labor) of milk production is higher in the northern region, while with family labor the farmers in the western region have higher profits.

Figure 7.4 Mean profit (including family labor) per liter of milk across different farm size groups and regions (daily milk production in liters)

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

However, higher profitability on small farms does not mean that the smallholder producers will not be displaced by large producers, because large farmers can compensate for lower per unit returns with large volumes, or they might be more efficient than small-scale producers.

Hypothesis 2: Small-scale producers are not less efficient if family labor is not costed and environmental externalities taken into account.

To test this hypothesis, we investigated the farm-level efficiency of Indian dairy farmers by estimating their technical and allocative efficiency. Unlike profits per unit, "efficiency" is a concept where the most efficient producers will drive out the least efficient ones over time. If smallholder farmers are more "efficient" in a nominal sense, the way to keep them involved in milk-production activity is to help them become better organized to enhance their "trust and reputation" and to acquire market power in buying inputs and selling outputs. If this is true, demonstrating it will help overcome the pessimism most policy makers and professionals have about the smallholder sector. The results of maximum likelihood estimates for the parameters of the stochastic profit frontier model and inefficiency model are discussed in the following section. Table 7.1 summarizes the variables used in the analysis.

Table 7.1 Definition and summary statistics of variables used in the model: All farmers

Variables	Unit	Mean	Minimum	Maximum	Standard deviation
Profit	Rs./liter milk	1.37	-0.48	0.35	1.155
Price of output	Rs./lit.	9.84	5.00	16.00	2.00
Price of fodder (dry and green)	Rs./kg	0.72	0.28	1.37	0.194
Price of feed concentrate	Rs./kg	5.82	3.00	9.10	0.935
Yield	Lit./day	7.35	1.5	22	2.764
Family labor	Minutes	8.50	0	67	9.389
Wage rate	Rs./day	50.15	0	95.00	9.643
Land	Hectares	0.08	0	0.76	0.102
Buildings and equipment	Rs./liter milk	0.25	0	1.92	0.234
Age of decision maker	Years	46	21	83	11.615
Education	Years of Schooling	7.40	0	17	4.755
Distance from market	Km	6.26	0.5	24.00	2.83
Membership in organization	Dummy	0.34	0	1	0.475
Access to information	Dummy	0.56	0	1	-
Access to credit	Dummy	0.33	0	1	0.471

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

The ML estimates for the parameters of stochastic frontier and inefficiency model for small, medium, large, and pooled categories are given in Tables 7.2 through 7.5 for the small, medium, and large farms, respectively for the pooled data. As expected, the signs of slope coefficients of the stochastic profit frontier with respect to price of output are positive and statistically significant and are highest (1.83) on small farms, followed by medium (1.62) and large farms (0.97). This indicates that if the price of milk were to be increased by 1 percent, the profitability would increase by 1.83 percent on small farms and 0.97 percent on large farms. The estimates for the parameters of the profit function with respect to fodder and feed were negative and statistically significant in all the cases. The negative elasticity for feed and fodder implies that profits tend to decrease as the expense of feed and fodder increases for the sample farms. The coefficient of yield-a proxy for technological change-is positive and statistically significant in all cases, which shows that with an increase in productivity, profitability also increases. The value of elasticity for family labor was positive and significant on small farms and negative but statistically insignificant on large farms. The wage rate had a negative effect on farm profitability but was significant only for medium and large farms. The coefficients of land and value of buildings and equipment do not appear to have any significant impact on small and medium farms, while on large farms both coefficients were positive and statistically significant.

Table 7.2 Parameter estimates of stochastic profit frontier function and inefficiency effects model for small farms: Pooled (n = 200)

Parameter	Coefficient	Standard error	t-value
Stochastic Profit Frontier Model
Intercept (b0)	-3.7117***	0.9221	-4.0254
Price of milk (b1)	1.8266***	0.1762	10.3655
Price of fodder (b2)	-0.5946***	0.1139	-5.2189
Price of feed (b3)	-0.4091*	0.2156	-1.8978
Yield (b4)	0.8738***	0.0956	9.1444
Family labor (b5)	0.3157	0.4645	0.6796
Wage rate (b6)	-0.2404	0.2604	-0.9233
Family labor x wage rate (b7)	-0.0834	0.2387	-0.3491
Land (b8)	-0.0245	0.0356	-0.6881
Buildings and equipment (b9)	-0.0276	0.0515	-0.5361
Land x Buildings and equipment (b10)	-0.0191	0.0333	-0.5739
Inefficiency Effect Model
Intercept (d0)	-3.9495**	1.9130	-2.0646
Age (d1)	0.6166	0.4979	1.2384
Education (d2)	-0.1558	0.1405	-1.1088
Distance from market (d3)	-0.0523	0.4995	-0.1047
Access to information (d4)	-5.0416***	1.0916	-4.6184
Access to credit (d5)	-1.0867*	0.5846	-1.8587
Environmental cost (d6)	-0.4619	0.2919	-1.5825
Zone dummy (d7)	3.5495***	0.7990	4.4427
Sigma-square (s2)	1.1133***	0.2047	5.4381
g (g = s2m /s2n)	0.9314***	0.0184	50.7040
Log likelihood ratio	-102.5239

Note: ^***, ^**, ^*: Significant at 1, 5, and 10 percent level, respectively.

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Table 7.3 Parameter estimates of stochastic profit frontier function and inefficiency effects model for medium farms: Pooled (n = 148)

Paramete1r	Coefficient	Standard error	t-value
Stochastic Profit Frontier Model
Intercept (b0)	-1.0172	0.9338	-1.0894
Price of milk (b1)	1.6231***	0.1340	12.1090
Price of fodder (b2)	-0.4034***	0.0764	-5.2797
Price of feed (b3)	-0.2726**	0.1357	-2.0082
Yield (b4)	0.2859***	0.0850	3.3652
Family labor (b5)	0.2781*	0.1577	1.7640
Wage rate (b6)	-0.6128***	0.1968	-3.1134
Family labor x wage rate (b7)	-0.1281	0.0801	-1.6001
Land (b8)	0.0442	0.0473	0.9336
Buildings and equipment (b9)	0.0834	0.0735	1.1354
Land x Buildings and equipment (b10)	0.0462	0.0501	0.9228
Inefficiency Effect Model
Intercept (d0)	0.8666	0.9735	0.8902
Age (d1)	-0.1515	0.2290	-0.6617
Education (d2)	-0.1291*	0.0679	-1.9020
Distance from market (d3)	0.2645	0.1617	1.6360
Access to information (d4)	-0.6850***	0.1636	-4.1876
Access to credit (d5)	-0.2624	0.1723	-1.5225
Environmental cost (d6)	0.7810***	0.1846	4.2316
Zone dummy (d7)	0.4525***	0.1706	2.6532
Sigma-square (s2)	0.0691***	0.0174	3.9703
g (g = s2m /s2n)	0.4524***	0.1756	2.5768
Log likelihood ratio	23.9918

Note: ^***, ^**, ^*: Significant at 1, 5, and 10 percent level, respectively.

Table 7.4 Parameter estimates of stochastic profit frontier function and inefficiency effects model for large and commercial farms: Pooled (n = 172)

Parameter	Coefficient	Standard error	t-value
Stochastic Profit Frontier Model
Intercept (b0)	-0.6711	0.8533	-0.7865
Price of milk (b1)	0.9672***	0.1885	5.1303
Price of fodder (b2)	-0.5244***	0.1341	-3.9100
Price of feed (b3)	-0.3270*	0.1759	-1.8594
Yield (b4)	0.3557***	0.0795	4.4724
Family labor (b5)	-0.0030	0.1349	-0.0224
Wage rate (b6)	-0.3903*	0.2305	-1.6931
Family labor x wage rate (b7)	0.0091	0.0663	0.1378
Land (b8)	0.0450***	0.0138	3.2702
Buildings and equipment (b9)	0.1333***	0.0432	3.0840
Land x Buildings and equipment (b10)	0.0359**	0.0164	2.1837
Inefficiency Effect Model
Intercept (d0)	-0.7041	1.8039	-0.3903
Age (d1)	0.0165	0.3494	0.0473
Education (d2)	0.1817	0.1324	1.3720
Distance from market (d3)	0.0388	0.5213	0.0745
Access to information (d4)	0.0810	0.1319	0.6138
Access to credit (d5)	0.2405	0.3340	0.7199
Environmental cost (d6)	-0.1155	0.1131	-1.0212
Zone dummy (d7)	-0.2891	0.3863	-0.7483
Sigma-square (s2)	0.0503*	0.0282	1.7828
g (g = s2m /s2n)	0.1754	0.5294	0.3313
Log likelihood ratio	23.7967

Note: ^***, ^**, ^*: Significant at 1, 5, and 10 percent level, respectively.

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Table 7.5 Parameter estimates of stochastic profit frontier function and inefficiency effects model for pooled sample: Pooled (n = 520)

Parameter	Coefficient	Standard error	t-ratio
Stochastic Profit Frontier Model
Intercept (b0)	0.9815	0.9862	0.9952
Price of milk (b1)	0.7952***	0.2259	3.5203
Price of fodder (b2)	0.3747***	0.1331	2.8155
Price of feed (b3)	-0.3493	0.2437	-1.4336
Yield (b4)	0.1686	0.1045	1.6133
Family labor (b5)	0.0718	0.1467	0.4896
Wage rate (b6)	-0.4014**	0.2028	-1.9788
Family labor x wage rate (b7)	-0.0143	0.0712	-0.2003
Land (b8)	0.0316	0.0304	1.0389
Buildings and equipment (b9)	-0.0386	0.0563	-0.6862
Land x Buildings and equipment (b10)	-0.0034	0.0370	-0.0914
Inefficiency Effects Model
Intercept (d0)	-10.7645***	1.6609	-6.4813
Age (d1)	2.7917***	0.4002	6.9752
Education (d2)	1.1659***	0.1047	11.1394
Distance from market (d3)	-0.5044***	0.1434	-3.5168
Access to information (d4)	-2.0976***	0.2601	-8.0633
Access to credit (d5)	-6.1533***	0.7950	-7.7396
Number of programs attended (d6)	-2.6338**	1.0896	-2.4172
Environmental cost (d7)	-0.6538**	0.2603	-2.5121
s2	1.0970***	0.1716	6.3922
gdmn	0.8969***	0.0281	31.9189
Log likelihood function	-125.8789

Note: ^***, ^**, ^*: Significant at 1, 5, and 10 percent level, respectively.

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Table 7.7 Parameter estimates of stochastic profit frontier function and inefficiency effects model for (small + medium + large farms): Western region (n = 260)

Parameter	Coefficient	Standard error	t-ratio
Stochastic Profit Frontier Model
Intercept (b0)	-3.6172***	0.9769	-3.7026
Price of milk (b1)	1.7897***	0.1509	11.8600
Price of fodder (b2)	-0.8051***	0.1285	-6.2647
Price of feed (b3)	-0.3943***	0.1492	-2.6429
Yield (b4)	0.6564***	0.0866	7.5784
Family labor (b5)	-0.5485**	0.2198	-2.4955
Wage rate (b6)	-0.1008	0.2271	-0.4439
Family labor x wage rate (b7)	0.3514***	0.1155	3.0411
Land (b8)	0.0907***	0.0234	3.8786
Buildings and equipment (b9)	0.0820*	0.0493	1.6650
Land x Buildings and equipment (b10)	0.0477**	0.0224	2.1340
Inefficiency Effects Model
Intercept (d0)	-5.0163**	2.1379	-2.3463
Age (d1)	1.5504**	0.6301	2.4605
Education (d2)	-0.2858	0.1794	-1.5932
Distance from market (d3)	-0.6360	0.4646	-1.3688
Access to information (d4)	0.5157	0.4333	1.1903
Access to credit (d5)	-1.1903***	0.4178	-2.8492
Environmental cost (d6)	1.2866***	0.4475	2.8750
s2	0.7104***	0.1300	5.4645
gdmn	0.8187***	0.0469	17.4484
Log likelihood function	141.08343

Note: ^***, ^**, ^*: Significant at 1, 5, and 10 percent level, respectively.

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Comparing the results of parameter estimates of the northern and western regions showed that the price of milk, feed and fodder, and yield had a significant impact on profit. The coefficients of variables associated with land and buildings and equipment were positive and significant in the western region but insignificant in the northern region. The coefficient of family labor had a perverse behavior in the western region.

The coefficients of explanatory variables in the inefficiency model of small and medium farms had the expected signs and were statistically significant in most cases (Tables 7.2 and 7.3), while in the case of large farms, none of the coefficients were significant (Table 7.4). Small and medium farmers with poor access to information and credit tend to be farther from the frontier and thus have lower technical efficiency, presumably due to low bargaining power. Further, small farmers, who spend more on environmental pollution abatement, tend to be closer to the frontier and technically more efficient because they consider manure a by-product and try to maximize their returns from manure. In contrast, medium and large farms consider manure disposal a liability and try to spend less on pollution abatement.

The coefficient of age in the efficiency model was positive and statistically significant in both regions. Older farmers tend to have lower levels of technical efficiency. The parameter estimate of access to credit was a significant determinant of inefficiency in both regions (Tables 7.6 and 7.7).

Finally, the parameter g, defined by g = s²_m /s²_n) is estimated to be close to unity for small farms, which suggests that the technical inefficiency effects are significant in the stochastic frontier model and that the traditional production function, with no technical inefficiency effects, is not an adequate representation of the data. The value of g was 0.45 for medium and 0.17 for large farms.

7.2 Technical Efficiencies of Dairy Farmers

The mean technical efficiency across different farm size groups is presented in Figures 7.5 and 7.6. The figures show that small farmers (<10 liters of milk) had the maximum efficiency (0.85); efficiency declined with farm size. The mean technical efficiency was estimated to be 0.81. This implies that, on average, the dairy farmers in the study area were producing milk to about 81 percent of the potential (stochastic) frontier production levels, given the levels of inputs and technology in use. The average technical efficiency of the sample farmers in the northern and western regions was estimated to be 0.79 and 0.84, respectively (Figure 7.6). The technical efficiencies showed a declining trend in the northern region with farm size, while in the western region there were no significant differences across farm sizes.

There was considerable variation in individual technical efficiencies (Figure 7.7). About 57 percent of the sample farmers had technical efficiencies in the range of 0.8-0.9. About 22 percent of the farmers had technical efficiencies in the range of 0.7-0.8. The figure indicates that, although there were very high relative frequencies of technical efficiencies above 0.80, a few farmers (< 2%) were quite poor in their technical efficiency performance.

Figure 7.5 Mean farm efficiency by farm size groups (daily milk production in liters): Pooled data

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Figure 7.6 Mean farm efficiency by farm size groups (daily milk production in liters): Pooled data

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Figure 7.7 Distribution of technical efficiencies of dairy farmers

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

The proportion of farmers with high technical efficiency (> 0.90) was significantly higher (75%) in the western region than in the northern region (about 60%) (Figure 7.8). Fewer than 3 percent of farmers in the western region had < 0.70 technical efficiency, while the proportion of such farmers was quite high (about 18%) in the northern region.

The above results clearly show that small dairy farmers were more efficient than large farmers, and the difference was more pronounced in the northern region.

Hypothesis 3: Profits per unit of small-scale producers are more sensitive to transaction costs than are those of large-scale producers.

We interpret hypothesis 3 as saying that transaction costs-such as lack of access to assets (capital, credit, liquidity, infrastructure) or information (livestock knowledge, commercial experience, vet care, market information, knowledge of input quality, trust of buyers)-play a bigger role in explaining the technical inefficiency of the small farmer than they do for the large farmer. To test the hypothesis, we compared the differential inefficiency of small and large farmers. The results, presented in Table 7.8, clearly show that variables such as access to information and technology and access to credit (proxy for transaction costs) were statistically significant in small farms, indicating that these factors are more important in explaining inefficiencies in small farms. On large farms, none of these factors seem to play an important role in explaining differences in the efficiency levels. We accept our hypothesis that profits of small farmers are more sensitive to transaction costs than those of large farmers. The estimate of parameter for the environment pollution abatement variable was negative and statistically significant in small farms, which indicates that expenditure on pollution abatement does influence (positively) the profits and efficiency of small farms.

Figure 7.8 Percentage of sample dairy farmers with technical efficiencies within selected ranges by regions

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Table 7.8 Differential inefficiencies (negative means more efficient) of small and large dairy farms

Parameters

Small                          (£ 10 dairy animals) Coefficient

Large             (> 10 dairy animals) Coefficient

Constant	-1.76	-0.71
Age of head of household	1.27	n.s.
Education of head	n.s.	n.s.
Distance to the market	n.s.	n.s.
Access to information and technology	-1.38	n.s.
Access to credit	-0.94	n.s.
Environment pollution costs	-0.49	n.s.
Region 2 (dummy for Western region)	2.91	n.s.

Note: n.s. = not statistically different from zero at the 5 percent level. All non-zero coefficients shown are significant at 5 percent or better.

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Hypothesis 4: Smallholder producers generate a lower negative environmental externality per unit of output than do large-scale producers.

Due to time and resources constraints, it was difficult to get reliable technical information about environmental externalities created by different categories of farms. Moreover, the difficulty arises with respect to different kinds of environmental externalities (positive as well as negative). Positive ones are possible, especially for ruminant manure, which is used as organic manure on the fields and which improves crop productivity. Negative externalities can vary; for example, smallholders (landless) may pollute by dumping, while huge concentrations of animals (peri-urban and commercial) on very large farms may produce much worse problems of odors, flies, and disposal of manure than smallholders. A further problem is that externalities by definition involve interactions with outcomes of the farm in question, such as the following:

• How to link externalities captured by one farm to costs borne by others.
• How to attribute effects on the environment outside the farm to the actions of that farm in particular.
• How to attribute effects on the environment inside the farm to the actions of that farm in particular.
• How to consider externalities, even if we know them, in an empirical production context.
• How to monetize the value of not paying for pollution created.
• How to annualize the costs and benefits of resource management, which are multiyear streams, at any one time.
• How to handle the problem that externalities captured and profits made are mutually dependent on each other.

In this study, we have focused on the extent of externality internalized by the farms by investing in pollution abatement. We have included the value of manure, labor cost spent on collecting and disposing of manure (either spreading on fields or making dung cakes for fuel), expenses incurred on manure pits, and so forth as a measure of pollution abatement. So we can redefine the hypothesis as "controlling for all other explanations of why farmers invest a given amount in pollution abatement (production activity, animals kept, herd size, sales, regulations, location, density of animal production in the immediate region, etc.) small farmers put more (or less) effort/investment into pollution abatement than large farmers." This analysis will also give insights into the type of farms that pollute the most for a given type of production and may identify policy-relevant farm characteristics that promote sustainable behavior. The results of average expenditure on pollution abatement across different farm sizes (milk production per day in liters) and the two regions are presented in Figures 7.9 and 7.10.

It is evident from the figures that small farmers spend more (Rs. 0.51/liter of milk produced) on pollution abatement than do large farmers (Rs. 0.15/liter). However, there were slight regional variations in spending on pollution abatement (Figure 7.10). The farmers in the western region generally spent more on pollution abatement than their counterparts in the northern region. One of the reasons for this differential behavior might be the fact that average chemical fertilizer consumption is much higher in the northern region, and farmers in the northern region do not attach much importance to organic manure. The average expenditure on pollution abatement in the western region was 0.46 rupees per liter compared with 0.40 rupees per liter in the northern region, but the trend was the same in both regions.

Figure 7.9 Average expenses (rupees per liter of milk) on pollution abatement in different categories (milk production per day) of farms

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

Figure 7.10 Average expenses (rupees per liter of milk) on pollution abatement in different categories (milk production per day) of farms: Northern and western regions

Source: IIM/IFPRI India Dairy Field Survey, 2002-2003.

The above results clearly show that smallholders have higher profits per liter of milk than large-scale producers. However, smallholders could still be driven out of the market:

• Large farmers are more efficient than smallholders (regional variations: small more efficient in western region and large in northern region).
• Large farmers produce larger volumes.
• Smallholders have difficulty complying with SPS/quality standards.
• Smallholders have less access to world dairy markets.

In the northern region, large farmers receive higher prices than small farmers; while in the western region, average prices received by small and large farms are almost the same. This indicates that cooperatives benefit smallholders more than large producers.

When family labor is not costed, farm profit efficiency and farm size had an inverse relationship, which shows that small farmers are more efficient than large farmers. The efficiency level was higher in the western region than in the northern region. There was also an inverse relationship between average expenses on pollution abatement and farm size, and farmers in the western region generally spend more for pollution abatement than their counterparts in the northern region.

The price of milk and productivity are the key factors affecting (positively) farm profitability, while the prices of inputs had a negative effect on farm profitability. There were major differences across farms in efficiency that can be traced to differences in access to information and credit. Inefficiency effects matter greatly for smallholders but not for large producers. Age of head of household and pollution abatement costs were two other important factors influencing efficiency on farms.

Smallholders have higher profits and are more efficient, but food safety and quality (SPS) issues will be more important in an open market environment and will have an impact on smallholder producers.