Livestock Research for Rural Development 17 (6) 2005 Guidelines to authors LRRD News

Citation of this paper

Production function analysis for smallholder semi-subsistence and semi-commercial poultry production systems in three agro-ecological regions in Northern provinces of Vietnam

Dinh Xuan Tung and S Rasmussen*

National Institute of Animal Husbandry, Hanoi, Vietnam
dxtung168@hotmail.com
*The Royal Veterinary and Agricultural University, Box 1870 Frederiksberg, Copenhagen, Denmark
sr@kvl.dk


Abstract

A formal cross section survey of 360 smallholder poultry keeping farms located in three agro-ecological regions in Vietnam was conducted. Cobb-Douglas production functions were applied to analyse and compare semi-subsistence and semi-commercial smallholder poultry systems in three regions. The general assumption is that the poultry production output at the farm level depends on the number of birds, feed amount, labor amount, garden size, income level and veterinary costs. In this study statistical tests were conducted to analyze the differences between regions and production systems.

The results from the analysis of production functions shows that the coefficients of flock size, feed amount per bird, labour amount per bird, household income level, and veterinary costs were highly significant in different models. Garden size has a significant influence only among the poultry farmers in Midland region. Regardless of region and system, the results indicate that the farm poultry output was most responsive to the variable feed per bird. However, the influence differed between production systems and regions.

The major conclusion of the analysis is that there was a significant difference in the estimated production model between the three regions. However, within regions, there was no significant difference between the two production systems, except in the region Highland.

Key words: Cobb-Douglas function, poultry production systems, production region, production systems, semi-commercial, semi-subsistence, Vietnam


Introduction

Poultry production in Vietnam is characterized by small-scale farms, in which birds are fed with mainly farm-produced grain combined with scavenging. This low input/output practice has been a traditional component of small farms in almost all areas of the country.

There are two main poultry production systems in Vietnam, semi-subsistence and semi-commercial poultry systems (definitions and characteristics are presented in section 2.1). In semi-subsistence poultry systems, small flock of local breeds are left to scavenge in backyards and garden area and fed with local available feeds. The salient features of the semi-commercial poultry systems are characterized by larger flock size with local or improved breeds and supplementary feeding with either grain or concentrate feeds or both. Semi-commercial farmers may have a better position for applying new technologies, thus they are more technically efficient than semi-subsistence farmers. Apart from that, farmers residing in different ecological regions have various comparative advantages, which also affect their production.

Poultry production plays an important role in providing food and cash income for villagers. The cash income in turn helps to contribute to poverty alleviation. Most of the studies in the field of village poultry production have focussed on characterising the production. (Dessie 1996; Swatson, et al 2001; Tadelle, et al 2003; Missohou, et al 2002; Aganga, et al 2000; Muchadeyi, et al 2004; McAinsh 2004; ); on feeding (Gunaratne et al 1994; Minh 1999; Gunaratne 1999; Rashid, M. 2003, Minh et al 2004; Henuk and Dingle 2002); on social-economics (Chitukuro et al 1997); on disease (Aini 1999; Tu 2002).

In this paper we analyse and compare the two poultry production systems mentioned by estimating Cobb-Douglas production functions based on data from a survey carried out in 2003. Results from such a type of analysis have not been published before. To the best of our knowledge, only Cevrger et al (2003) used multiple regression models to estimate the impact of several factors affecting the profitability of broiler production in Turkey, and Ojo (2003) applied the stochastic frontier production function for analysing technical efficiency in Nigerian egg production.

The specific purpose of the research was to analyze and compare the production within the two poultry production systems in three ecological regions in Vietnam.

The paper is organized in the following sections. In section 2, main characteristics of two production poultry systems and three regions are described. Section 3 deals with research methods, which include field research design, sample selection, and data collection. Section 4 presents research hypotheses, which are developed based on the general understanding of the production systems in the three regions. In section 5 the method of data analysis is discussed and the results of the statistical analysis are presented and discussed. The final section includes some conclusions.

The major conclusion based on the analysis of 360 semi-subsistence and semi-commercial poultry farms in three ecological regions is that there is a significant difference in production technology between the three regions, but that the production systems did not show significant differences between the estimated Cobb-Douglas production function.


Main characteristics of production systems and ecological regions

Poultry production systems

Based on criteria of feeding systems and flock size, the poultry production systems in lowland, midland and highland regions can be categorized into two groups, including (i) Semi-subsistence poultry systems; and (ii) Semi-commercial poultry system.

Semi-subsistence poultry system

It is estimated that around 90% of smallholder poultry farms carry out poultry production as semi-subsistence production system (Oanh 2000). This system has existed for the long time in all parts of Vietnam, where poultry production is an integrated part of a diversified farm. A family may keep up to 50 birds, confined to mainly indigenous breeds. The most common breed is of local Ri breed, which is characterized by small bodyweight of 1.2-1.6 kg at 6 months of age; hens produce 10-15 eggs per clutch, equivalent to egg yield of from 45-60 eggs per hen per year, depending on management practice. Chicks are hatched from their hen eggs, but sometime farmers buy chicks from local markets to supplement their flock. Most of farmers keep poultry all year round.

The characteristic of this system is low inputs/low output. The birds are left to scavenge in the backyards or in the garden areas and the feed picked up in the garden area are supplemented with limited amount of home produced grain such as paddy rice or maize, and kitchen waste. The amount of feed given to birds does not focus on production efficiency, and depends heavily on the availability of grain that farmers have in their storages and eating needs from their birds. Commonly, birds are freed and fed (as a breakfast) in the morning and fed again in the evening and then locked in during the night in houses. Poultry housing is simple, it is made by bamboo, wooden with thatched or tile roof.

Little labour and investment are required for this system, since it is considered as a side-line activity, birds are not been given proper attention concerning disease prevention and treatment, and therefore the mortality rate is high. Normally, when disease occurs farmers go to the local drug store to buy medicine based on advice from the shopkeeper, and they treat poultry disease by themselves. Farmers who practice this system still work in crop or other activities for their main livelihood.

Farmers can sell different poultry products when they need cash, such as chicks, growers, broilers, hens, cocks, eggs or other poultry species like ducks. Selling products normally takes place in local markets or through by-pass traders (assemblers), or by sale directly to customers or other farmers.

Semi-commercial poultry farmers

This system is the combination of the traditional system and the improved technologies. Breeds kept in this system consist of specialized or a mixture of local and improved breed with the flock size ranging from 50 to 200 birds. Generally, chicks of improved breeds are bought from local hatcheries and local chicks are from local markets. The majority of semi-commercial farmers can keep a certain number of laying hens to produce chicks for fattening. In the case hat they do not have enough hens to produce the chicks as they need, they can exchange eggs to other farmers in order to get larger number of eggs to be hatched. By doing so, farmers have bigger flocks in certain periods. This type of collaboration is sometimes not only based on economic considerations, but also on social relationships. From hatching to one month of age, chicks are kept with their hens in cages to brood and protect them; older birds are allowed to scavenge in backyards or in the gardens at certain times during the day and brought back to their housing in the evening. The housing varies from permanent to simple houses, made by mainly local materials.

Besides the naturally available feed resources such as worms, insects, pest, vegetable, and grass that the birds can scavenge, they are also fed by using grains and/or commercial feeds that are bought from local feed shops. Although these farmers have the financial capacity to buy some concentrate feed, this system is still a part-time or supplementary generated income activity on these households in the three regions. Household members are still engaged in other farming activities like cropping, or other livestock or off-farm employments.

Measures on disease prevention, treatment and management are given higher consideration compared to those in semi-subsistence farmers, including vaccination. Chicks are paid more attention and given better feed rations.

Like semi-subsistence farmers, outputs of semi-commercial poultry farmers are eggs, live birds, which include chicks, growers, broilers, hens, cocks and other poultry species like ducks and geese. These outputs are either consumed by the households or are sold to different buyers like assemblers or wholesalers and consumers.

This production system may be a transition stage from the semi-subsistence to commercial poultry system. Farmers who are involved in this system mainly are former government employees, current local officers or wealthy farmers who have permanent income and skills, especially knowledge of market conditions. It seems that know-how and capital are important factors abreast for development of commercial poultry production.

General features of the regions

The three regions (Lowlands, Midlands, and Highlands) where data was collected are shortly described in the following. More detailed information can be found in Tung (2003).

Lowlands

Topography is almost universally flat, river delta land with an average altitude of 3-6 m above sea level. It is appropriate for irrigated cropping systems. Per capita land area is limited to 500 m2. It is the most densely populated region in the country, characterised by intensive rice production and pig husbandry. The major agricultural activity in the area is rice growing, utilizing a double rice cropping system intermixing with one cash crop in winter such as maize, bean, or vegetables. Livestock production is an integrated farming activity for the majority of the households. This region is located close to consumption centres, and limited land areas, combined with inadequate off-farm employment opportunities cause a large segment of youth immigrate into urban areas for employment. Inhabitants are mainly Kinh group. In general, infrastructure and the level of education in this area are good.

Midlands

The typology consists of a mix of relatively flat rice production land and clay soil foothills. Apart from rice as a main crop, other crops such as groundnut, bean, and sugar canes are also important. Cropping systems in this region are one or double rice crops on low laying land, and cassava or sugarcane on rolling soil hilly land. Per capita land area in the region is 3.5 times higher than that of their counterpart living in lowland region. Small livestock plays an important role in farming households. Besides poultry, pigs and cattle production are also an integrated farming activity. The inhabitants are mainly Kinh and Tho groups. Infrastructure and education level in this region are better than in the highland area

Highlands

This region is a mountainous area; the topography is of a mix of relatively terraced rice cropping land and limestone mountains. Crops grown in this region are mostly native varieties, characterized by low productivity. Land for rice plantation is very limited (around 80 m2 per person). Cassava is planted on hilly land, which is 500 m2 per capital and bamboo growing is performed on the mountainsides. The inhabitants are mainly ethnic minority groups such as Thai, Tho, Dao... Villager's livelihood depends on natural and planted forest like exploitation of timber. The infrastructure in the zone is very underdeveloped.

Poultry Inputs-Outputs clarification
Outputs

Both semi-subsistence and semi-commercial poultry farmers in the three regions keep poultry flocks as an integrated farming activity, consisting of hens, cocks, chicks, growers, broilers and other poultry species such as ducks and geese. The output of this type of production includes eggs and all of these birds' categories. Farmers can use eggs and broilers for both home consumption and sale, but when cash is needed, hens, cocks, chicks, growers, broilers and other poultry species such as ducks and geese are also sold to get money. These products are not homogenous.

Inputs

In order to involve in poultry production, apart from what birds can scavenge in backyard or in garden areas, farmers have to feed them with grains such as paddy rice, broken rice, maize, rice bran, cassava. These grains can be home produced or purchased from market. On some farms, especially semi-commercial farms, birds are also fed with purchased concentrate feed.

Sometime farmers have to buy feed, medicines or even birds for replacement. To do so requires capital which is provided by using their own cash or by taking up loans.

As mentioned previously, poultry relies partly on scavenging feeds in gardens areas. Consequently, the size of gardens, where the birds can search for food has an impact on feed cost.

To take care of birds including, feed preparation, feeding, watering, cleaning shelters and disease prevention and treatment, farmers have to spend time on these activities. In household farms, the households typically consist of three adults and two children, with the wife being the main responsible for poultry production. However, this proportion changes slightly for the group of semi-commercial poultry farms. The proportion of wives conducting poultry farming activities is lower, while the rate of husband activities is higher. Wives were most often involved in activities such as sale of products, feeding, watering, and cleaning the chicken house, while husbands are most often involved in other activities such as veterinary prevention and treatments.

According to the description above, output of household poultry enterprises depends on the number of birds on the farm, the amount of various feed inputs provided by the farmer, garden area where birds can search for food, the amount of labour spent on taking care of birds, capital, and veterinary input (measured in veterinary cost)..


Research hypotheses

According to the analysis in the previous section, we assume that the poultry production output at the farm level depends on the number of birds, feed amount, labor amount, garden size, income level, and veterinary costs.

We expect that the number of birds per farm and the feed amount is positively related to the amount of output. We assume that semi-subsistence and semi-commercial production systems not only rely on provided feed, but also on scavenging in household gardens. Garden size has positive relationship with production capacity. We also expect labour amount to be positively related to output, because more labour invested, results in more care, which should lead to higher output. Total household income may have positive or negative impact on poultry production capacity. Disease costs for poultry health care per farm per year, which includes medicines, administration costs may either be positively or negatively related to production.


Data and methods

Sample selection

The analysis was carried out based on survey data from 12 villages in the two provinces of Thai Binh and Thanh Hoa. The villages were purposively selected as representatives for their agro-ecological area. The four villages in Nam Trung commune, Thai Binh province were selected as representatives for lowlands, the four villages in Thanh Tien commune, as representatives for midlands and the other four village in Quan Hoa, Thanh Hoa province as representatives for Mountainous area.

Design of the study

The selection of households within the study area was done using stratified and systematic random sampling techniques. The survey covered a total of 360 households. 120 households were selected for data collection in each area. In order to ensure adequate representation of two groups of poultry keepers, the poultry producers in each village were stratified into two groups of semi-subsistence and semi-commercial poultry farms based on the two criteria of current flock size and feeding regime.

Semi-subsistence poultry farms: Households, whose family have 1-50 scavenging birds, which are given a very limited amount of feed besides what the birds get from scavenging.

Semi-scavenging poultry farms: Households, whose family have a flock of more than 50 and the birds are fed with grain and high quality feed combined with scavenging in garden areas.

Poultry farm groups were developed using the following method: In each village, a meeting was held with the village leader, one representative of the farmer association, one representative of Women's Union and a commune animal health worker to split poultry farms into the two groups based on the lists of households in each village. Finally, the poultry farms were systematically selected to be representatives for their own region and production systems.

Research methods

The information was collected using formal questionnaires for household interview. The interviews were conducted at household premises. Household members were explained clearly about the purpose of the study in order to provide as reliable information as possible. The interviewees were asked to provide information on a recall basic, so they had to rely on memories of household members. In many cases, both husband and wife were at the interviews, so that they could complement each other to give very detailed information about their farms.

A questionnaire assessing basic information at household level was designed in accordance with a set of indicators. The selection of indicators reflected the objectives of the study.

In order to assist in data collection, two assistants were selected and trained in each area. The assistants were selected locally as they understand well the local farming systems, culture and local languages, and they could conduct the interviews during the day or at night depending on when farmers are available. Households were given a small gift as incentive and compensation for the time dedicated to talking with the researchers.

The assistants were trained about objectives of the study, data collection and questionnaire procedures. After the training, assistants and author carried out pre-testing to check the appropriateness and relevance of the questionnaire in term of sequences and wording of questions. After pre-testing, the questionnaire was improved accordingly.

The field assistants with the help of the author carried out the actual surveys. Almost all interviews were conducted with the participation of more than one family member, so that each person could contribute with information relevant to the activities known to them. For example, heads of households normally would know more about the amount of land that they own, while his wife is better able to provide information on variable inputs related to animal husbandry.

Data analysis methods

In order to examine the above-mentioned hypotheses, the following model is applied:

Y = f(X1,X2, X3, X4,X5,X6) (1)

Where:

Y is output of poultry products;
X1 is number of birds
X2 is amount of feed
X3 is garden size
X4 is amount of labour
X5 is household income
X6 is veterinary costs
f( ) is a production function

All data are based on individual farm observations. The data was prepared as follows:

Per farm poultry output

As described in previous section, outputs of household poultry enterprises include a number of products, which can be used for home consumption or sold on the market. These products include eggs, chicks, growers, broilers, hens, cocks and other poultry species such as ducks or geese. In order to measure the output level, different products had to be aggregated into one measurement unit. We used two alternative methods for aggregating output data.

The first method is based on output-values. Output is simply measured by value. i.e. in Vietnamese currency unit.

where: 
Yi is output of i´th-farm (i = 1,…., 360), 
Wik is market price of ith-farm for product k, and 
Zik is the quantity of product k (including home consumption and/or sale) for the ith farm. 
The index k is a product index, including eggs, chicks, growers, broilers, hens, cocks, and other poultry species products.

This method of aggregating is simple and easy to apply, because the quantity of the different categories of products used for home consumption or for market sale and their respected prices were collected.

The second method of aggregating production data was based on the Laspeyere's quantity index:

where: 
Yi is output of ith-farm (i = 1,…, 360), 
Wk0 is the price that the base farm received for poultry product k, 
Zki is quantity of product k for the ith farm, and 
Zk0 is quantity of product k for the base farm.

Per farm inputs

Number of birds per farm (X1)

The different types of birds were converted into "chicken unit" by using the following formula:

Where: 
X1i is number of "birds" raised on the ith-farm and 
Wk is average price of the k'th-category of birds (chicks, growers, broilers, cocks, hens and other), 
Xik is number of birds in category k on farm i (chicks, growers, broilers, cocks, hens and other), and 
Wb is average price of broilers.

Amount of feeds (X2)

Feed used for household poultry enterprises is also includes a number of different feed items, consisting of paddy, broken rice, maize, cassava, concentrate and other. To measure feed input, we use two alternative methods to aggregate the different feed items:

In the first method, the total feed amount is measured in energy units (Kcal). We assumed that the quality of each type of feed is similar for all sample farms. All feed types used for poultry are converted into Energy unit (Kcal) by multiplying the quantity of each type of feed used with its calorie content.

In the second method, the total feed amount is measured in monetary values (thousand of Vietnamese currency units). All feed types used for poultry are valued at market prices. We also assumed that the average price for each type of feeds is similar for all sample farms in the same region.

Garden size (X3)

Garden area is measured in m2 per farm. We use garden size as a one input, since in household poultry production, birds are left to scavenge in garden to search for food besides what they are fed by the farmers.

Amount of labour (X4)

The labour is measured in person-hours. In both poultry production systems, mainly wives performed almost all poultry farming activities including feed preparation, feeding, watering, and shelter cleaning to poultry sale. However, sometime husbands or children also participated.

Household income (X5)

Total household income (thousands VND/farm/year), was used as a proxy of capital input. In order to maintain and develop the poultry enterprise, some investment is required.

Disease costs (X6)

Disease costs for poultry health care per farm per year, which includes medicines and administration costs.

A summary of data from the 360 farms is shown in Table 1

Table 1: Statistical description of variables used in the model

Factors

Mean value

Standard deviation

Max value

Min value

Output, 000.VND/farm/year

Lowland:

 

- Semi-subsistence farms

1801.0

1508.8

5350.0

80.0

- Semi-commercial farms

5083.4

2538.6

10872.0

1030.0

Midland:

 

- Semi-subsistence farms

1676.0

1039.1

7779.4

54.0

- Semi-commercial farms

8521.4

8923.1

44632.9

1454.6

Highland:

 

- Semi-subsistence farms

710.6

369.7

1869.4

0.0

- Semi-commercial farms

3103.8

3518.5

17663.4

910.5

Number of poultry/farm

Lowland:

 

- Semi-subsistence farms

16.8

13.8

64.6

1.3

- Semi-commercial farms

56.0

43.5

140.8

10.6

Midland:

 

- Semi-subsistence farms

25.0

15.1

70.3

5.3

- Semi-commercial farms

103.5

129.9

738.2

8.7

Highland:

 

- Semi-subsistence farms

13.9

9.6

47.7

1.3

- Semi-commercial farms

57.2

46.7

243.1

6.8

Feed costs, 000.VND/farm/year

Lowland:

 

- Semi-subsistence farms

965.6

835.4

4191.2

51.0

- Semi-commercial farms

2895.2

2147.2

9880.0

411.2

Midland:

 

- Semi-subsistence farms

753.8

464.0

2992.0

8.8

- Semi-commercial farms

7533.9

16127.7

82737.2

1178.2

Highland:

 

- Semi-subsistence farms

501.7

270.1

1307.8

44.8

- Semi-commercial farms

1748.6

1964.5

8089.0

281.5

Garden area, m2/farm

Lowland:

 

- Semi-subsistence farms

334.5

317.6

1440.0

0.0

- Semi-commercial farms

442.3

354.9

1368.0

108.0

Midland:

 

- Semi-subsistence farms

628.4

570.7

3500.0

0.0

- Semi-commercial farms

917.6

730.4

2500.0

0.0

Highland:

 

- Semi-subsistence farms

464.0

443.6

2500.0

0.0

- Semi-commercial farms

721.8

918.1

5000.0

0.0

Labour, man-hours/farm/year

Lowland:

 

- Semi-subsistence farms

150.9

90.3

480.0

15.6

- Semi-commercial farms

282.5

212.3

1100.0

75.4

Midland:

 

- Semi-subsistence farms

97.1

92.9

838.0

5.05

- Semi-commercial farms

410.4

336.2

1260.0

34.7

Highland:

 

- Semi-subsistence farms

165.7

134.6

765.0

19.0

- Semi-commercial farms

361.0

429.1

2388.0

40.6

Household income, 000.VND/farm/year

Lowland:

 

- Semi-subsistence farms

6130

4230

19310

900

- Semi-commercial farms

12420

5520

29230

4350

Midland:

 

- Semi-subsistence farms

7140

4400

25880

910

- Semi-commercial farms

13390

6610

30090

4330

Highland:

 

- Semi-subsistence farms

6400

4280

21880

980

- Semi-commercial farms

1376

6990

28840

4100

Veterinary costs, 000.VND/farm/year

Lowland:

 

- Semi-subsistence farms

22.8

56.7

502.0

0.0

- Semi-commercial farms

48.8

46.0

185.0

0.0

Midland:

 

- Semi-subsistence farms

40.6

41.1

190.0

0.0

- Semi-commercial farms

188.0

141.2

544.0

0.0

Highland:

 

- Semi-subsistence farms

9.2

18.6

100.0

0.0

- Semi-commercial farms

46.6

      79.6

314.0

0.0

After all input and output data were prepared, all independent and dependent variables were re-scaled around the value of 1 by dividing each variable by the average value.

Parameter estimation and tests

Function specification

A Cobb-Douglas function was chosen as the functional form of the production function. The reason for choosing this type of production function is that it is linear in its logarithmic form, and therefore easy to estimate by using ordinary least squares estimation technique (OLS). At the same time, this function type has been widely used for production function analysis by many authors (Beringer 1956; Heady 1951; Seyoum et al 1998). The function has the following form.

Where: 

Yi is poultry output of the i'th-farm 
X1i is Number of birds
X2i is Feed amount
X3i is Garden size
X4i is Labour amount
X5i is Total household income
X6i is Disease costs
and a, β1, …, β6 are parameters to be estimated and ei is an error term.

Prior to the model estimation, the variables were examined for multi-co-linearity. We found that there are high correlations (coefficient of correlation, r>0.75) between X1 and X2 and between X1 and X4. In order to reduce estimation problems, we divided X2 by X1 and X4 by X1 to create two new variables: X21 (feed amount per bird) and X41 (labour per bird).

The revised model is therefore:

Taking logarithms on both sides, we get:

ln(Yi )= ln(a) +β*1lnX1i + β2lnX21i+ β3lnX3i+ β4lnX41i+ β5lnX5i + β6lnX6i + ei

The parameters β*1, β2, β3, β4, β5, and β6 were estimated using Ordinary Least Squares analysis (OLS) and after estimation the value of β1 was estimated as,

The hypothesis is that there may be differences both between production systems and between regions. To account for these potential differences, and to establish an appropriate econometric model for testing these differences, dummy variables were added for both production systems and regions as follows:

D = dummy variable for production systems. D = 1 for semi-subsistence and D = 0 for semi-commercial production system.

R2 = dummy variable for regions: R2 = 1 for midland, and R2 = 0 for otherwise, 

R3 = dummy variable for highland: R3 = 1 for highland, and R3 = 0 for otherwise

The complete model therefore has the following form.

lnYitr     = (α 0 + α 1D + α2R2+ α3R3) + (β*1+ β*11D+ β*12R2 + β*13R3) lnX1

 + (β2+ β21D+ β22R2 + β23R3) lnX21 + (β3+ β31D+ β32R2 + β33R3) lnX3

 + (β4+ β41D+ β42R2 + β43R3) lnX41 + (β5+ β51D+ β52R2 + β53R3) lnX5

+( β6+ β61D+ β62R2 + β63R3) lnX6   + eitr                                

where t refers to production system and r refers to region.

The model assumes a single additive effect of either production system or region on each parameter. However, the model does not include any interaction between region and production system. Such an interaction may exist, but is not tested here, because of the limited data set.


Findings

Estimated parameters of the full model

The model in (4) was estimated using Ordinary Least Squares Analysis (OLS) All estimations and tests were performed using the procedure REG in SAS ver. 8.02 ((SAS/STAT Guide for Personal Computers Version 6 Edition). The estimated parameters are shown in Table 1A in the Appendix.

The R-sq value is 0.76 indicating that the independent variables included in the production function explained about 76% of the variations in the production of poultry of both groups of semi-subsistence and semi-commercial poultry farms.

The parameters β*1, β2, β3, β4, β5, and β6 are the production elasticities for the semi-commercial production system in the Lowland region, i.e. the relative change in production divided by relative change in the amount of the input in question. Elasticities for the other production system and regions are estimated by adding the parameters of the relevant dummy variables. All the elasticities for the semi-commercial production system in the Lowland region have a positive sign as expected except β3 (which is not statistically different from zero). The production elasticities are as follows: Number of birds: 0.73. Feed per bird: 0.63. Garden size: 0. Labor per bird: 0.14. Household income: 0.28. Disease cost: 0.11. Thus, for instance, the production increases by 7.3 % when the number of birds increases by 10 % (conditional on unchanged feed per bird and unchanged labor per bird) and by 6.3% when the amount of feed per bird is increased by 10%.

The coefficients of the dummy variables both have positive and negative signs. To analyze the possible differences between production systems and regions, a number of tests described in the following were performed. All the tests were carried out as F-tests on a 5 % test level. (Griffiths et al 1993).

Test of full model

The first test is the following (here and in the following, the term H0 refers to the hypothesis to be tested against the alternative hypothesis H1)

Test 1.  Whether there are no differences, neither between production systems nor  regions?

Ho:

All regions and systems are equal

α 1 = β*11 =  β21 =  β31 =  β41= β51= β61 2 = α 3 = β*12 = β*13 = β22 = β23 =   β32 = β33= β42 = β43= β525362 = β63=0

H1:

At least one system or region is different

α 1 or  β*11 or β21 or  β31 or  β41or β51or β61 α 2 or α 3 or β*12 or β*13 or β22 or β23 orβ32 or β33or β42 or β43or β52 orβ53 orβ62 or β63 0

The F-test rejects H0 (accepts H1). The F-value is 1.76, compared to an F-value of 1.55 at the 5% significance level. Therefore, at least one of the regions or one of the production systems is statistically different from the others when based on the total data set. Therefore the following two tests were performed:

Test 2.  Whether there is a difference between production systems?

Ho:

All systems are equal

α 1 = β*11 =  β21 =  β31 =  β41= β51= β61 = 0

H1:

At least one system is different

α 1 or  β*11 or β21 or  β31 or  β41or β51or β61 0


Test 3. 
Whether there are differences between regions?

Ho:

All regions are equal

α 2 = α 3 = β*12 = β*13 = β22 = β23 = β32 = β33= β42 = β43= β525362 = β63=0

H1:

At least one region is different

at least one of the above ≠ 0

In Test 2, H0 is not rejected (F-value =0.53 compared to an F-value of 2.01 at the 5% significance level). Thus, together with the result of Test 1, this result reveals that the regions are statistically different. This was formally confirmed in Test 3, in which H0 was formally rejected (F-value is =2.10 compared to an F- value of 1.67 at the 5% test level)

Estimated model for each region with and without dummy variable

Because Test 3 reveals that the regions are different, individual estimations of parameters were performed for each of the three regions. The detailed estimation results are shown in Table 4A, 5A and 6A in the Appendix

Test of regional models

The final tests performed include tests of whether there are significant differences between production systems within regions. Thus, for each region, the following test was performed:

Test 4.  Whether there is difference between production systems within a region? (Estimate for each region)

Ho:

All systems are equal within a region

α 1 = β*11 =  β21 =  β31 =  β41= β51= β61 = 0

H1:

At least one system is different

α 1 or  β*11 or β21 or  β31 or  β41or β51or β61 0

For both Lowlands and Midland regions, the H0 was not rejected. Therefore, the final models for each of these two regions were estimated without the dummy variable for production systems. The detailed results are shown in Table 3A. The final model for the two regions is the following:

For households in the Lowlands, the final estimated regression equation is:

LnY = -0.67784 +0.80929lnX1 + 0.57341lnX21 -0.02486lnX3 +0.22342lnX41 +0.21154lnX5 + 0.14116lnX6

All independent variables had statistically significant effect on the production at the 5% test level, except garden size (lnX3).

Of the significant variables, the variables for number of birds had the strongest effect on the production, followed by feed per birds, household income and disease costs.

For households in the Midlands, the estimated regression equation is:

LnY = -0.7118 + 0.55121lnX1 + 0.31232lnX21 -0.08592lnX3 + 0.03580lnX41 + 0.19224lnX5 + 0.15894lnX6

Almost all variables had statistically significant production elasticities, except labor costs and garden size, where the elasticities were very low.

Similar to the lowland model, the elasticity of production was highest for number of birds (0.55), followed by feed per bird (0.31), but their values were lower than those in the Lowland model.

However, the test for the Highland region had as the result that H0 is rejected, and that therefore there is statistical difference between the two production systems within Highland region. The final estimated functions for the Highland region are (see table 6A in the Appendix):

Semi-subsistence poultry system in Highlands

lnY = - 1.38116 + 0.42393lnX1 +0.09573lnX21 + 0.02335lnX3 +0.22564lnX41+0.24422 lnX5 +0.1627 lnX6

The elasticity of production with respect to number of birds and feed amount per birds was 0.10 and 0.42, respectively, and were significant at the 1% test level. The coefficient of the variable of feed per bird (X21) was the smallest compared to the same variable in other regions or production system.

Semi-commercial poultry system in Highlands

lnY = -0.65963 + 0.69465lnX1 +0.77443lnX21 +0.05916lnX3 -0.06946lnX41+0.33823lnX5 +0.15364lnX6

For semi-commercial farms, the coefficient for number of birds was 0.69. Feed per bird variable remained the important contributor to the production. The implication is that holding other variables constant, a 10% increase in feed amount per bird in semi-commercial farms would result in 7.7% in production output, while it is only 0.9% in semi-subsistence farms.

In the Highland region, the four variables such as garden size, labor per bird, household income, and disease costs were non-significant, indicating that there is no clear relation between these inputs and the production. Variables for feed amount per bird had the strongest effect on the production in semi-commercial farms, but it loses its role in semi-subsistence farms.

By regions, the variable for number of birds had the strongest effect on the production in all model, except the highland semi-commercial model, the highest elasticity was found in the lowland model (0.81) followed by the highland semi-commercial model (0.69), the midland model (0.55) and the highland semi-subsistence model (0.42).

The variable for feed per bird had the second largest elasticity in all models, except the highland semi-commercial model. The production increased by 5.7% for the lowland, 3.1% for the midland, 0.9% for highland semi-subsistence and 7.7% for the highland semi-commercial model when the amount of feed per bird is increased by 10%.

Compared with the highland models, the coefficient of the garden size turned out to be negative in the lowland and midland models, while labor per bird had positive sign in all model, except the highland semi-commercial model. The effect of household income and disease costs on the production had positive sign.

Comparison of aggregation methods

As we mentioned in section 4.4, poultry production output per farm was aggregated by using two different methods, i.e. by value (Y-v) and by formal index (Y-i). Feed input was also aggregated using two different methods, i.e. by value and by energy units. Thus, a total of 2×2 = 4 different data sets were available.

The results presented in the previous sections refer to the data with output measured by index (Y-i) and feed input measured by costs. Estimation was also carried out for the other three combinations. Comparisons show that there were only small differences between the different aggregation methods.

Limitations of the study

Smallholder poultry farming is mainly based on scavenging with small flock of different ages, which makes it difficult to get reliable data. Feed supplements to the birds are also ad hoc making it difficult to get reliable information on feed intake. The sale and purchase of poultry products are also ad hoc. It is therefore difficult to get authentic data on the marketing of these products.

In addition, because of time and fund constraints, the study covered only twelve villages and a limited number of households were interviewed. Although the data are considered to be representative of smallholder chicken farming systems of the Northern provinces of Vietnam as a whole, caution has to be applied when using the results of the study in other regions.


Discussion

The variables for number of birds and feed per bird are positive and statistically significant in all regions. This indicates that these variables have significant impact on the smallholder poultry production. The positive and strongest impact of the number of birds on the production (except the highland semi-commercial model) showed that the smallholder poultry production is largely extensive, and consistent with the nature of the smallholder poultry production, where the output is dependent largely on number of birds. The variable feed per bird remained the second most important contributor to the production, except in the highland semi-commercial model. This shows that supplementary feed has a high production effect in smallholder poultry production, especially in the highland semi-commercial farms. Minh (1999) concluded that 22 % to 32 % of hen nutrition requirement is derived from scavenging.

The results presented here suggest that the farm poultry output could be increased by simultaneously increasing the number of birds and the feed amount per bird.

Garden size did not show a significant impact on output in any of regions. This reveals that perhaps in both production systems, birds are fed mainly on household supplementary feed rather than searching for food in gardens as we thought. It may be true, because garden land area per bird for semi-subsistence farms is 16 m2, and it is 6.5 m2 for semi-commercial farms. These areas are limited for birds to search feed daily. The coefficient of garden size turned out to be negative in the lowlands and midlands compared with the highland. The reasons for this difference could be the fact that garden cropping systems in the lowlands and midlands are more intensive - with annual cash crops - than that in the highlands - with perennial trees. This could lead to a reduction in the number of birds scavenging in the gardens, where cash crops are planted.

Labor per bird had positive sign, except the highland semi-commercial model. This parameter is not significantly different from zero in any models, except the lowlands. Although labor input has a low influence on production, it is dangerous to make any definite conclusions concerning the productivity of labor. Further analyses are needed.

Household income level did have significant effect on poultry production in all regions, except highlands. This indicates that capital is an important factor for development of poultry in the lowland and midland regions.

Disease costs were also found to be insignificant in all regions, except the midlands. This result indicates that farmers could increase the poultry production if proper animal health care is provided.


Conclusions


References

Aganga A A, Omphile U J, Malope P, Chabanga C H, Motsamai G M and Motsumi L G 2000 Traditional poultry production and commercial broiler alternatives for small-holder farmers in Botswana. Livestock Research for Rural Development. 12(4): 1-8 http://www.cipav.org.co/lrrd/lrrd12/4/Aga124a.htm

Aini I 1999 Diseases in rural family chickens in south-east Asia. First INFPD/FAO Electronic conference on family poultry, The scope and effect of family poultry research and development (Guèye E F Editor) http://www.fao.org/ag/againfo/subjects/en/infpd/documents/econf_scope/paper2.html

Beringer C 1956 Estimating enterprise production functions from input-output data on multiple enterprise farms. University of Idaho. Journal of Farm economics.Volume 38, pp: 923-930.

Cevger Y and Yalcin C 2003 A quantitative model to determine factors affecting profits of broiler enterprises. Turk Journal of Veterinary and Animal Science, Volume 27, pp.1201-1205.

Capalbo S M and Antle J M 1988 Agricultural Productivity: Measurement and Explanation. Resources for the Future. John Hopkins University Press, Balimore.

Chitukuno H R and Foster H A 1997 Methodology for enabling women to improve poultry productivity through better husbandry and disease control. Proceedings of an International Workshop held on June 13-16, 1995 at the International Livestock Research Institute, Addis Ababa, Ethiopia

Dessie T 1996 Study on village poultry production systems in the central highland of Ethiopia. MSc. Thesis. Swedish University of Agricultural Science, Uppsala, Sweden.

Griffiths W E, Hill RC and Judge G G 1993 Learning and practicing econometrics. John Willey and Sons, INC.

Gunaratne S P 1999 Feeding and nutrition of scavenging village chickens. Electronic conference on family poultry, The scope and effect of family poultry research and development (Guèye E F Editor) http://www.fao.org/ag/againfo/subjects/en/infpd/documents/econf_scope/add_paper2.html

Gunaratne S P, Wickramaratne S H G, Chandrasiri  A D N and Roberts J A  1994 Scavenging village chicken production and the scavenging feed resource base in areas in Sri Lanka with distinct wet and dry seasons. Final Report of the ACIAR Small Project in Sri Lanka on Scavenging Village Chickens. Veterinary Research Institute and James Cook University.

Heady E O 1951 A production function and marginal rate of substitution in the utilization of feed resources by dairy cows. Iowa State University. Journal of Farm economics.Volume 33, pp: 485-498.

Henuk YL, Dingle JG 2002 Practical and economic advantages of choice feeding systems for laying poultry. World's Poultry Science Journal. 2002; 58(2): 199-208

McAinsh CV, Kusina J, Madsen J and Nyoni O 2004. Traditional chicken production in Zimbabwe. World'-s-Poultry-Science-Journal. 60(2): 233-246.

Minh D V 1999 Effect of energy and protein on the production performance of local and improved hens in northern Vietnam. MSc thesis, Swedish University of Agricultural Sciences, Uppsala, Sweden.

Minh D V, Lindberg J E and Ogle B 2004 Effect of scavenging and protein supplement on the Feed Intake and Performance of Improved pullets and Laying hens in Northern Vietnam. Asian-Aust. J. Anim.Sci. Vol 17, No. 11: 1553-1561

Missohou A, Dieye P N and Talaki E 2003 Rural poultry production and productivity in southern Senegal. Livestock-Research-for-Rural-Development. 2002; 14(2): 1-6. http://www.cipav.org.co/lrrd/lrrd14/2/miss142.htm

Muchadeyi F C, Sibanda S, Kusina NT, Kusina J and Makuza S 2004. The village chicken production systems in Rushinga district in Zimbabwe. Livestock Research for Rural Development. http://www.cipav.org.co/lrrd/lrrd16/6/much16040.htm

Oanh N T 2000 Feed Industry in Vietnam. Department of Agriculture, Ministry of Agriculture and Rural Development, Vietnam

Ojo S O 2003 Productivity and technical efficiency of poultry egg production in Nigeria. Department of Agricultural Economics and Extension, Federal University of Technology, 704 Akure, Nigeria.

Rashid M 2003 Nutritional status of scavenging chickens with special emphasis on energy and protein supplementation under rural condition in Bangladesh. MSc. Thesis, Royal Veterinary and Agricultural University, Denmark.

Seyoum E T, Battese G E and Fleming E M 1998 Technical efficiency and productivity of maize producers in eastern Etheopia: A study of farmers within and outside the Sasakawa-Global 2000 project. Agricultural Economics, Volume 19. pp. 341-348.

Swatson HK, Nsahlai IV and Byebwa B 2001 The status of smallholder poultry production in the Alfred district of Kwazulu-Natal, South Africa: Priorities for intervention. Proceedings of the 10th Conference of the Association of Institutions for Tropical Veterinary Medicine, Copenhagen, Denmark 2001

Tadelle D, Million T, Alemu Y and Peters K J 2003 Village chicken production systems in Ethiopia: Flock characteristics and performance. Livestock Research for Rural Development. 15(1): 1-9. http://www.cipav.org.co/lrrd/lrrd15/1/tadeb151.htm

Tu T D 2002 Village poultry production dominates in Vietnam. World-Poultry. 18 (7): 12-15

Tung D X 2003 Instututional settings for small-scale poultry production in Thai Binh and Thanh Hoa provinces. Small Livestock Component-Agricultural Sector Program Support, Hanoi, Vietnam.



Appendix

Table 1A.  Estimated coefficients of Cobb-Douglass Production function for sample poultry keeping farmers involving in semi-subsistence and semi-commercial production systems residing in 3 regions

Variable

Parameter

Coefficient Estimate

Standard 
Error

t Value

Pr >|t|

Intercept

α0

-0.58512

0.10092

-5.8

<.0001

lnX1

β*1

0.73419

0.12971

5.66

<.0001

lnX21

β2

0.63043

0.11449

5.51

<.0001

lnX3

β3

-0.0006

0.09531

-0.01

0.995

lnX41

β4

0.13615

0.13722

0.99

0.3222

lnX5

β5

0.2833

0.14574

1.94

0.0532

lnX6

β6

0.1077

0.07695

1.4

0.163

Dt

α1

-0.20316

0.1064

-1.91

0.0575

DlnX1

β*11

0.05445

0.12915

0.42

0.6737

DlnX21

β21

-0.10001

0.11258

-0.89

0.3753

DlnX3

β31

-0.02272

0.0881

-0.26

0.7967

DlnX41

β41

0.15176

0.11724

1.29

0.1969

DlnX5

β51

-0.1114

0.14291

-0.78

0.4365

DlnX6

β61

0.01159

0.07033

0.16

0.8693

R2

α2

-0.00766

0.10572

-0.07

0.9423

R3

α3

-0.24953

0.12865

-1.94

0.0537

R2lnX1

β*12

-0.30812

0.13314

-2.31

0.0216

R3lnX1

β*13

-0.14953

0.137

-1.09

0.2763

R2lnX21

β22

-0.28044

0.12528

-2.24

0.0262

R3lnX21

β23

-0.2293

0.14063

-1.63

0.1044

R2lnX3

β32

-0.0762

0.0897

-0.85

0.3965

R3lnX3

β33

0.09398

0.10978

0.86

0.3928

R2lnX41

β42

-0.23712

0.14892

-1.59

0.1127

R3lnX41

β43

-0.04805

0.14778

-0.33

0.7454

R2lnX5

β52

-0.00291

0.13151

-0.02

0.9824

R3lnX5

β53

-0.0161

0.14196

-0.11

0.9098

R2lnX6

β62

0.02863

0.08593

0.33

0.7393

R3lnX6

β63

0.08099

0.08469

0.96

0.34

Root MSE=0.48038.     R-Square = 0.76

.

Table 2A: Estimated coefficients of Cobb-Douglass Production function for sample farmers involving in semi-subsistence and semi-commercial poultry production systems across regions

Variable

Parameter

Parameter Estimate

Standard Error

t Value

Pr>|t|

Intercept

α0

-0.64957

0.08337

-7.79

<.0001

lnX1

β*1

0.54739*

0.09603

5.7

<.0001

lnX21

β2

0.47439

0.08558

5.54

<.0001

lnX3

β3

-0.0301

0.07359

-0.41

0.6829

lnX41

β4

-0.03849

0.09198

-0.42

0.676

lnX5

β5

0.17682

0.12593

1.4

0.1616

lnX6

β6

0.15631

0.05772

2.71

0.0073

Dt

α1

-0.10075

0.10651

-0.95

0.3452

DlnX1

β*11

0.12514

0.12313

1.02

0.3105

DlnX21

β21

-0.01526

0.10774

-0.14

0.8875

DlnX3

β31

-0.0288

0.08595

-0.34

0.7379

DlnX41

β41

0.09141

0.11226

0.81

0.4163

DlnX5

β51

-0.07541

0.14156

-0.53

0.5947

DlnX6

β61

0.04867

0.06893

0.71

0.4808

Root MSE =0.49915.        R-Square = 0.7287

Table 3A: Estimated coefficients of Cobb-Douglass Production function for sample poultry keeping farmers residing in three regions (refer to test 3 in the test plan)

Variable

Parameter

Parameter Estimate

Standard Error

t Value

Pr > |t|

Intercept

α0

-0.67784

0.07256

-9.34

<.0001

lnX1

β*1

0.80929*

0.08898

9.1

<.0001

lnX21

β2

0.57341

0.0821

6.98

<.0001

lnX3

β3

-0.02486

0.06968

-0.36

0.7216

lnX41

β4

0.22342

0.11477

1.95

0.0528

lnX5

β5

0.21154

0.08135

2.6

0.0099

lnX6

β6

0.14116

0.05675

2.49

0.0136

R2

α2

-0.03405

0.0971

-0.35

0.7261

R3

α3

-0.26105

0.12144

-2.15

0.0326

R2lnX1

β*12

-0.25808

0.1217

-2.12

0.035

R3lnX1

β*13

-0.15285

0.13525

-1.13

0.2596

R2lnX21

β22

-0.2611

0.12146

-2.15

0.0326

R3lnX21

β23

-0.26097

0.13911

-1.88

0.0619

R2lnX3

β32

-0.06107

0.08885

-0.69

0.4926

R3lnX3

β33

0.11374

0.10772

1.06

0.2922

R2lnX41

β42

-0.18762

0.14517

-1.29

0.1975

R3lnX41

β43

-0.01413

0.14524

-0.1

0.9226

R2lnX5

β52

-0.0193

0.13034

-0.15

0.8824

R3lnX5

β53

-0.00192

0.13929

-0.01

0.989

R2lnX6

β62

0.01778

0.08251

0.22

0.8296

R3lnX6

β63

0.06056

0.0812

0.75

0.4566

Root MSE = 0.48098.    R-Square =  0.7555

 

Table 4A: Estimated coefficients of Cobb-Douglass Production function for sample lowland farmers involving in semi-subsistence and semi-commercial poultry production systems

Variable

Parameter

Parameter Estimate

Standard Error

t Value

Pr>|t|

Intercept

α0

-0.58462

0.16151

-3.62

0.0005

lnX1

β*1

0.41003*

0.22357

1.83

0.0708

lnX21

β2

0.31145

0.18151

1.72

0.0905

lnX3

β3

-0.12313

0.15441

-0.8

0.4278

lnX41

β4

-0.06515

0.20558

-0.32

0.7522

lnX5

β5

0.03355

0.25741

0.13

0.8967

lnX6

β6

0.06652

0.12338

0.54

0.5914

Dt

α1

-0.03324

0.19431

-0.17

0.8647

DlnX1

β*11

0.47727

0.25327

1.88

0.0635

DlnX21

β21

0.36929

0.2088

1.77

0.0812

DlnX3

β31

0.10462

0.17552

0.6

0.553

DlnX41

β41

0.34226

0.25458

1.34

0.183

DlnX5

β51

0.14572

0.27431

0.53

0.5969

DlnX6

β61

0.12201

0.14361

0.85

0.3984

Root MSE =  0.49047.     R-Square = 0.7543

Table 5A: Estimated coefficients of Cobb-Douglass Production function for sample midland farmers involving in semi-subsistence and semi-commercial poultry production systems

Variable

Parameter

Parameter Estimate

Standard Error

T Value

Pr > |t|

Intercept

α0

-0.51752

0.21452

-2.41

0.0180

lnX1

β*1

0.36850*

0.16469

2.24

0.0279

lnX21

β2

0.37843

0.13571

2.79

0.0066

lnX3

β3

-0.28491

0.13727

-2.08

0.041

lnX41

β4

-0.09769

0.16253

-0.6

0.5494

lnX5

β5

0.20118

0.22116

0.91

0.3656

lnX6

β6

0.21788

0.12551

1.74

0.0863

Dt

α1

-0.38618

0.24544

-1.57

0.1194

DlnX1

β*11

-0.03457

0.22412

-0.15

0.8778

DlnX21

β21

-0.21222

0.18688

-1.14

0.2594

DlnX3

β31

0.23407

0.15066

1.55

0.1241

DlnX41

β41

0.09681

0.19677

0.49

0.624

DlnX5

β51

-0.01548

0.24869

-0.06

0.9505

DlnX6

β61

-0.09889

0.14439

-0.68

0.4953

Root MSE =  0.47156 .     R-Square =  0.7055

Table -6A: Estimated coefficients of Cobb-Douglass Production function for sample highland farmers involving in semi-subsistence and semi-commercial poultry production systems

Variable

Parameter

Parameter Estimate

Standard Error

t Value

Pr > |t|

Intercept

α0

-0.65963

0.17357

-3.8

0.0004

lnX1

β*1

0.69465*

0.22251

3.12

0.0029

lnX21

β2

0.77443

0.18985

4.08

0.0001

lnX3

β3

0.05916

0.2024

0.29

0.7712

lnX41

β4

-0.06946

0.14665

-0.47

0.6376

lnX5

β5

0.33823

0.26925

1.26

0.2144

lnX6

β6

0.15364

0.10832

1.42

0.1617

Dt

α1

-0.72853

0.24445

-2.98

0.0043

DlnX1

β*11

-0.27072

0.2617

-1.03

0.3054

DlnX21

β21

-0.6787

0.2307

-2.94

0.0048

DlnX3

β31

-0.03581

0.21911

-0.16

0.8708

DlnX41

β41

0.2951

0.17998

1.64

0.1068

DlnX5

β51

-0.09401

0.29983

-0.31

0.755

DlnX6

β61

0.00906

0.12733

0.07

0.9435

Root MSE  =  0.42583.      R-Square  = 0.8249


Received 20 December 2004; Accepted 6 January 2005; Published 1 June 2005

Go to top