V. Data and Survey

This study utilizes data from 2 main sources: (a) the secondary data collected by various government agencies and surveys reported in previous research, (b) a socio-economic survey of 480 livestock farms. Moreover, the researchers also conducted interviews with the executives of the agro-business firms and some large-scale farmers. The first sources of data are used for the analysis of the livestock revolution and description of the livestock sector. The farm survey is designed for an estimation of the profit frontier faced by the livestock farms, taking into account the impact of technological change and environmental abatement on the scaling up process. So the survey contains not only the cost and revenue of the farms, but also the cost of environmental abatement, food safety, and proxies of transaction costs that affect the farms size.

This chapter will describe how the livestock farm survey was carried out. The following sections will be discussed in order: field research design, sampling methods, design of the questionnaires and pre-test of questionnaires, and problems faced by the researchers.

5.1 Field Research Design and Site Selection

Since our research objectives are to quantify the impact of the technological change, market forces, externalities, and government policies on the scale of livestock production, particularly the small holders, the survey has to cover both the old major livestock-producing areas and the new production areas that have been affected by the livestock revolution in the last 15 - 20 years. The coverage of new growth areas will allow us to assess the impact of an expansion of the modern livestock production on the small holders who practice the traditional method of production. The second important criterion is that those areas must be the main location of four kinds of livestock, i.e., broiler, layer, swine, and dairy. There are a number of livestock growing areas that meet our objectives. For example, the existing largest livestock areas are in the Eastern and the Western provinces (see Table 5.1). The new areas include two provinces in the East, one in the West, a few provinces in the North and the Northeast.

This study decided to choose a cluster of 6 provinces with connecting borders in 3 regions, namely Chachoengsao, Cholburi, and Srakaew in the East; Saraburi, and Lopburi in the Central and Korat in the Northeast (see Figure 5.1). There are two justifications for our location decision. First, these provinces are the largest producing areas for broiler, layer and dairy and one province, Chacherngsao, is also the second largest swine province after a Western province of Ratchaburi. Chachoengsao is the largest producing areas of broiler and layer; while Korat and Saraburi are the main dairy producing provinces. In recent years, the dairy farms have expanded farther east to Srakaew and to the Central province of Lopburi. New layer farms can also be found in Nakorn Nayok (which is not in our sample), and the connecting province of Saraburi and Lopburi. There are also new broiler ad swine farms in both provinces. Korat is both the old location for dairy farms and swine farms which are relatively small. But the new dairy farms are also expanded into other eastern districts of Korat. The second reason for such a cluster of sample farms is to economize on both the survey expenses and survey time. In addition, all of these provinces are less than 250 kilometers from Bangkok, which make it very convenient to effectively manage the survey.

5.2 Sampling Methods

After considering the number of population farms in six provinces and the budget constraint, the researchers have determined the appropriate sample size for each type of livestock farms, 486 farms in total, with a consultation from the FAO consultants. The number of the planned sample size is reported in Table 5.2. Note that the sample size for the layer and dairy farms are the smallest, but 85 farms are large enough for the required econometric work.^[96]

Given the objective of assessing the scaling-up impact of the livestock revolution, the researchers decided to classify the livestock farms into two broad class size, large-scale and small-scale as follows:

The classification serves only as a convenient guide to select sample farms, and does not intend to convey a message that it represents the actual farm size in Thailand. In fact, the modern large-scale farms for broiler and swine are much larger than the cut-off criteria shown above.

The next step is to allocate the sample size into the sampled provinces, using the provincial share in the total number of individual livestock subsector production within the sampled provinces. For swine, broiler and layer, the sample farms come from 5 provinces of Cholburi, Chachoengsao, Saraburi, Lopburi and Korat. Srakaew is excluded because it has very small number of those livestock farms. For dairy, Chachoengsao is excluded for the same reason.

Within each province, the sample size for each kind of livestock farm is allocated into two-class size, using the individual subsection size distribution in each province. The data comes from the Department of Livestock Development. The result of the planned sample's breakdown is given in Table 5.3.

The survey was carried out between November 2002 and January 2003. In the survey, the surveyors were asked to increase the sample size by 5 percent so those samples with inadequate information can be dropped without any significant impact on the sample size. The actual numbers of sample farms are shown in Table 5.3.

Two approaches were used to choose the sample farms. The first approach is to obtain farm address from the informed sources, i.e., Department of Livestock Development, the dairy and layer producer cooperatives, the producer associations, the university professors who are the consultant to the farmers, and four agri-business companies (CP, Betagro, Panus Pokaphand, and Bangkok Feed Mill). Then the farm owners were approached by phone to set up an appointment. Most of the swine sample farms were obtained through this approach. Therefore, the sample farms are affected by the farm owners' willingness to participate in the survey. The second approach is the random sample. The surveyors would obtain the information about the main location of livestock farms from the local authorities and then drove the car around the areas, scouting for the farms that were willing to cooperate. Most of the sample farms in dairy, broiler and layer were accessed by this approach. The weakness of our method of sample selection is discussed later.

5.3 Designing the Questionnaires and the Pretest

There are 4 types of questionnaires one for each livestock product, i.e., broiler, swine, layer, and dairy. In order to capture all the needed information, the questionnaires are divided into 6 main board areas:

• Farms Characteristics

• Cost of production of livestock commodities (both variable and fixed cost)

• Sale of livestock commodities

  (1) Farms Management
  (2) Environmental Management
  (3) Characteristics of the farm owners

The questionnaires had been pre-tested in 6 occasions (started from 1/10/02) in the following provinces Chachoengsao, Cholburi, Lopburi, and Nakhonpathom. Out of these 6 occasions, we had completed 20 samples of questionnaires, five samples for swine, five dairy cows, three for broilers and seven samples for layers.

As the result of pretest questionnaires, we faced a significant number of problems, which get solved in our groups meetings, in total of 7 times. The problems that we faced are as followed. Firstly, we did have to adjust our questionnaires eight times mainly due to the misunderstanding of farming techniques and miscommunication between the interviewers and interviewees.

However, there are still major obstacles to overcome. The problem was getting in contact with the owners of the farm for the interview because the majority of the farms (especially swine, broilers, and layers) do not trust outsiders with several reasons. Firstly, the fear of disease infection brought about by outsiders, so we tried to overcome this problems by interview them outside of their farms. Secondly, most farmers, particularly the commercial large and medium-scale, are afraid of a few things, e.g., the tax authority, hiring of cheap illegal workers, and over-using of antibiotic. Although the problems could not be entirely overcome, the researchers have tried to use our personal connection with the feed companies, large agribusiness companies (CP and Betago) and the Department of livestock in order to gain their trust and to get into contact with these farmers. Lastly, due to the insufficient in information in Thailand, we faced massive problem of finding the name and the address numbers of the farms. We tried to overcome this problem using various data sources. For example, personal connection, feed companies, cooperative, farmers organization, Department of Livestock, commercial banks, and private companies. But the sources are still ineffective, as some organizations did not want to help us due to commercial reasons.

5.4 Selection and Orientation of the Interviewers

The surveyor was divided into 3 teams, each consisting of 7-8 surveyors and one supervisor. The first team, which consisted of college students, was responsible for the survey in the broiler and layer farms in Cholburi and Chachoengsao. The second team of college graduates carried out the surveys of the swine, dairy, broiler, and layer farms in Korat, Saraburi, and Lopburi. Finally the last team, also the graduate students, did the survey of swine farms in Cholburi and Chachoengsao.

In order to get ready for the survey, we had trained all surveyors in three occasions one in Cholburi (27/11/02), Korat (01/12/02), and Bangkok (8/12/02). After the training, each surveyor had to carry out two experimental interviews under the supervision of the TDRI research supervisors. The purposes were twofold, i.e., to make sure that the surveyor fully understood the questionnaire and to acquaint than with the questionnaire before the actual survey. Each individual surveyor was trained to be able to do all the 4 types of questionnaires.

5.5 Quality of Data and Problems with the Survey

In all spectrums of fieldwork, there will always be some external and inconceivable problems regardless of the input efforts to perfect the work; this survey too was not spared. Throughout the data collection period, we have encountered many difficulties. In this section, we will discuss some of the major obstacles and problems in our survey, their origins, their direct effects, and any further impacts. This should help clarifying some of the limitations that our collected data may have. Nevertheless, we too have taken precautions and strategies to minimize these errors.

5.5.1 Time Constraints

Limited time span has always been the most eminent enemy for fieldwork. The given span of time for the survey was rather short, allowing no cushion for mishaps. This, in turn, further pushed all surveyors to work faster in a rush, and to compromise for what at hands to keep up with the time constraint. Geographical locations of farms often also aggravated the pressure, as many farms are located sparsely not as a cluster, some even into an almost hermit area of the provinces. The teams must spend extra efforts and time in searching out for their precise locations. For example, when the appointed time for interviews could not be met as planned due to the above difficulties, the teams must compromise for a simple substituting sample via random sampling around the area. Consequently, the initial plan for controlling samples according to farm size must also be compromised (compare the actual and plan sample size in Table 5.3).

5.5.2 Cooperation from Interviewees

With random sampling of farms, the most cooperative farms were those of small to medium scale, while on the contrary, the mid-large to large farms refused to participate without prior appointment. This also affected the stratified control of sample size initially planned. For those with proper appointment, large farms would merely give vague answers and figures, sometimes underestimates especially concerning the sales and profits, in fear of taxes. For smaller size, applying both for appointment and random sampling, their willingness to cooperate was truly applaudable, though they largely lacked systematic information records. It is worth a note that corporate farms refused altogether to cooperate and to participate with the study. Thus information obtained during the surveying period is purely consisted of data collection from non-corporate farmers. Moreover, some contract farms under the wings of big corporations prohibiting any release of information also hesitate to reveal anything. Some do not cooperate as they are instructed by their contractor not to give any interview without prior permission from the company. The researchers managed to get permission from the large-scale firms to interview a small number of these contract farms.

5.5.3 Questionnaire's Limitation

For each type of questionnaire, it took around 1-2 hour per farm to complete. The form also contains many repetitive questions aiming at the same information, one with details and the other just an estimated ballpark figure. The purpose for such a perplexing construction was for verification of the information validity. However, they also have adverse effects, as repetition could easily lead to losing interests and boredom. Furthermore, in order to respond to some questions on the form effectively, the interviewees must be keeping a good standard of systematic record, which appeared as not the case. Many farm owners, especially small-scale ones, responded rather from memory than hardcopy records. As noted earlier, larger farms that maintained well-organized records have chosen to give vague answers, instead of specific and to the point.

5.5.4 Language Barriers

Like it sounds, there were also communication problems between the surveyors and the surveyees, arising primarily from the use of technical terms used on the form. While the locals used different terms, even in native words, what appeared on the forms is always very technical and scientific. Therefore, the barriers induced miscommunication of the true objectives of the questions asked.

5.5.5 Specific Problems to Individual Livestock Subsector

Swine. Most large farms have reported sales and profits lower than the actual, evidently when compared to the structures and the size of their farms, and calculations that showed constant business lost over the years. For smaller farms, the organization and systematic records remained the central problem. Moreover, farmers have also reported numbers from their senses, not proper estimation. As for the amount of feeds, some used current figure as the past average, while some did the vice versa.

Broiler. We found no apparent problems specifically pertaining to this subsector. However, as it is dominated by contract farming, Our survey teams found it extremely difficult to stratified data as initially wanted.

Layer. Large layer chicken farms have their secret ratio for mixing the ingredients for layer's feeds. Thus, the calculation of the feed costs turned into nightmare with all hidden information. The other problem is similar to those of broiler's, the lack of management variation. Though in this case, independence dominates the subsector.

Dairy cow. This subsector also has big difficulties in the cost of feed calculation, especially with raw feeds (green grass). Some farms simply let cattle graze on the ranches, and some farmers seek their own grass elsewhere. Thus, when we asked for estimates, the estimations often appear higher than the average given by technical benchmark.

All of these, admittedly, have reduced the quality of our data set. Fortunately, we have taken many measures to abate these problems in our data set. Getting perfect information was not the feasible objective, as all experienced researchers would expect, though we wanted to ensure that we would get as close as possible to the best set. After the data collection and the fieldwork, we brought together all that we have collected and inspect to mitigate errors during one of the longest periods of the project, the data editing period.

We have trained all of our surveyors to record down all possible figures on the paper, while the data editing team would keep records of the sources of all information. This built a very complex and detailed database with a trace back system to the source of any errors or inconsistencies. For example, the two approaches to the same figure, detailed and ballpark figures, as described above, were utilized during this period. Before it worked against our data quality, ironically it also worked in many ways for it. We have used both approaches as a comparison for consistency, and if some inconsistencies were found, our teams would contact directly back to and would verify with the farms that have provided such information. For example, we contacted back up to 30 farmer to validate some information in the swine subsector. To reinforce our inspection even further, we also brought in some objective standards used among the livestock exports to form a benchmark for checking the consistency of our data. Therefore, though the data set might be imperfect, it is a valid one with enough quality to produce robust conclusion. In addition, for broiler subsector that is currently dominated by contract farming, figures and numbers are systematically kept and could be traced back far enough in the past. Thus they produce only few to none inconsistencies, and are very trustworthy. For dairy cattle subsector, cooperatives have provided great influence in maintaining correct and verifiable records, particularly individual farm records of milk sale and feed purchase through the cooperative. Both of these industries have filled in the weak point of our set, and have bolstered its validity.

Table 5.1 Livestock Production by Regions and Provinces, 1987 - 2000

Source: Swine, Layer and Dairy Cow from Department of Livestock Development, 2000 Broiler from Center for Agricultural Information, Office of Agricultural Economics, 2000

Note: Growth rate for year 1996-2000 ** Growth rate for year 1992-2000

Figure 5.1 Location of Livestock Sample Farms

Largest production areas

• Chachoengsao: swine, broiler and layer

• Saraburi & Korat: dairy

Table 5.2 Number of Sample Farms by Animal

Table 5.3 Sample Size by Province and Type of Livestock Farm (No. of farms)

Source: TDRI, see text.