| Source of technology and funding | IFAD-supported research carried out by the International Livestock Research Institute (ILRI) |
| Expected Benefit: | Improved efficiency in delivery of livestock health services to smallholder livestock keepers, leading to increased use of services and improved livelihoods based on livestock-keeping |
| Crops and enterprise: | Livestock-keeping |
| Agro-ecological zones: | All |
| Target region and countries: | Developing countries |
| Keywords: | Farmer preferences, livestock health services, delivery, disease control, conjoint analysis |
Introduction
Low uptake of livestock health technologies among smallholder livestock keepers is often attributed to the poor provision of veterinary services. In some cases, this is related to lack of infrastructure and trained personnel; however, in many cases, livestock keepers face high non-price (transaction) costs that constrain their use of available services. These costs are reflected in the criteria which form farmer preferences for service provision. Understanding farmer preferences can therefore help identify those service characteristics that, if improved, would potentially enhance farmer uptake. This note describes a practical approach for the assessment of farmer preferences for livestock health services that can be used to identify opportunities for the improvement of the efficiency and effectiveness of service provision.
When a consumer purchases a good or a service, a choice is often made among a range of competing products. To make the choice, the consumer may consider several different criteria such as price, brand, colour and size. The final choice reflects the consumer’s preferences with respect to these criteria. Similarly, livestock keepers evaluate several different factors, whether consciously or subconsciously, when deciding to use a given livestock health service or product. For livestock keepers in developing countries, the attributes are likely to include, besides price, those that influence the substantial non-price costs incurred to procure and use the service or product. An obvious example is the proximity of the service provider; especially in an area where service providers are widely dispersed, the farmer will want to minimize the transport and time costs required to seek a service provider and so is more likely to prefer service providers who are nearby. Poor infrastructure and low density of service provision make this type of search cost much more significant in developing countries than it is in developed countries and therefore more likely to play an important role in farmer decision-making.
One approach, then, for improving farmer access to and use of livestock health technologies is, first, to identify the attributes of these technologies and of provision that are most important in farmer decision-making and then design technologies or interventions that respond more closely to these attributes.
A technique can be borrowed from commercial marketing studies to assess farmer preferences with respect to livestock health service provision. The technique is called ‘conjoint analysis’ and is the standard method used by marketing professionals to test potential consumer demand for new products. The technique involves characterizing a product by its key attributes and then asking consumers to rank a set of candidate products representing selected combinations of different attribute levels. For toilet soap, for example, the attributes may include price, size, colour, fragrance and packaging. Each attribute may have two to four levels, such as red, green, or white for colour. This may result in hundreds of potential combinations as candidate products. Statistical methods are used to identify the minimum subset of combinations needed to sort out the trade-offs that consumers make between individual attribute levels when they do the ranking. The rankings are analysed using regression analysis, which generates a relative score for each individual attribute level. Based on the scores, one could answer the question, ‘how preferable is white soap over red?’. More importantly, the analysis provides estimates of the relative importance of the different attributes in the consumer’s ranking decision. For example, the soap fragrance may be estimated to account for 40% of the consumer’s decision, whereas colour may only account for 10%.
The following section outlines the steps for the application of this technique to livestock health service provision. Illustrations are taken from a study of veterinary service provision so that smallholder dairy farmers could confront a tick-borne cattle disease in Kenya.
Applying the conjoint analysis technique
1. Identify a commonly used livestock health service. To evaluate farmer preferences, it is necessary first to formulate a well-defined context. This should relate to a common preventive or curative service that is familiar to farmers. In the Kenyan example, farmers were asked to consider what they would do to seek treatment for a dairy cow suspected of suffering from East Coast fever, an often fatal tick-borne disease.
2. Identify key attributes of the service provision. What three to six key factors are farmers likely to consider when deciding to use a given service provider, particularly in the context chosen in #1? These factors may be selected based on expert opinion or on interviews with key informants or farmers themselves. For the Kenyan study, we first identified the various types of transaction costs and risks associated with finding and using a service provider. We found that, when grouped together, these costs and risks suggested a number of candidate attributes. Risk of a service provider not performing as expected or failing to honour a service agreement is minimized, for instance, if the farmer has worked with the provider before or knows farmers who have. Hence, level of previous experience with a service provider was identified as an important attribute, in addition to five others: price, distance (to the provider’s office), level of qualification, promptness of service and the offer of bundled services (additional features such as credit, free advice on other problems, etc.).
3. Formulate two or three levels for each attribute. The levels selected should cover the full possible range whether actual or expected, while remaining feasible. We considered, for example, a low, medium and high value for price and two levels of promptness: quite quick (within 48 hours) and quite slow (only within six days).
4. Generate an orthogonal set of combinations. SPSS? has a specific command (Data/Orthogonal Design/Generate) in the Categories module (versions 6-8 for Windows) that allows one to enter the attributes and their levels, and it will provide a minimum set of combinations for the ranking exercise. The minimum number of combinations required increases quickly as the numbers of attributes and their levels grows, so it is best to keep the numbers low.
5. Design illustrations of the attribute levels and represent the combinations on laminated cards. The illustrations must be able to convey effectively the attribute concept to avoid confusing the farmer when asking him to rank the different combinations. Each card represents a single combination of one attribute level for each of the attributes. The farmer ranks the combinations by physically ordering the cards. Two such cards from the Kenya study are shown in the box.
6.
Carefully train the enumerators. It is very important that the
enumerators administering the conjoint experiment be well trained and
that they present a consistent introduction describing the choices and
instructions to all farmers. The instructions must be sufficiently pre-tested
to ensure that they do not confuse the farmer. Emphasis must be placed
on avoiding undue influence on the farmer’s rankings and biasing
the results.
7. Administer the ranking experiment to a sample of the target farmers. Any standard sampling strategy for a typical survey can also be used for the selection of farmers for the conjoint experiment. The enumerator introduces the experiment and asks the farmers physically to rank the cards in order of preference. The enumerator then records the ranking. It is useful at the same time to collect additional information about the farmer’s characteristics (gender, experience, farm size, etc.) so as to understand the factors correlated with different preference patterns.
8. Analyse the rankings. A regression procedure described in standard conjoint analysis references is used to analyse the rankings. SPSS? has a user-friendly Conjoint command (also found in the Categories module of versions 6-8) that automatically performs all of the necessary transformations and re-scaling and directly generates the final outputs. For each individual farmer, these outputs consist of estimated utility (part-worth) scores for each attribute level, relative importance scores for each attribute, and goodness-of-fit measures. Average values are also computed for the full sample. The table below reports the sample averages from the Kenya study:
| Attribute and attribute levels | Part-worth score | Relative importance score |
| 1. Price Low price Medium price High price |
0.71 -1.42 -2.13 |
9.1% |
| 2. Professionalism Veterinarian Paraveterinarian Agro-vet storekeeper |
1.39 0.28 -1.66 |
20.3% |
| 3. Reputation Have heard nothing Heard from other farmers Know the provider |
-0.77 0.10 0.67 |
10.3% |
| 4. Distance | 0.94 -0.94 |
11.2% |
| 5. Promptness of service Within 2 days Only within 6 days |
3.20 -3.20 |
37.4% |
| 6. Bundled services Credit, free advice given No credit or free advice |
0.97 -0.97 |
11.6% |
| N=284 smallholder dairy farmers in Kenya Central Highlands. | ||
The part-worth scores are unitless and so are interpreted as relative values. A positive score represents a positive preference, with higher values indicating stronger preferences. A negative score indicates attribute levels that farmers have a preference against. For example, price is seen to be a negative attribute at all three levels considered and increasingly negative as the price increases, as might be expected. Among service providers, veterinarians are preferred (value of 1.39) over paraveterinarians (0.28), who in turn are much preferred over storekeepers (score of -1.66). The highest positive preference among attributes is for fast service, which exhibits a score of 3.20.
The relative importance of each attribute is measured as a percentage by the ‘relative importance score’. For Kenyan smallholder dairy farmers, promptness of service is clearly the most important attribute, contributing 37% on average of the farmers’ decisions when choosing among different service provision options. Professional qualifications follow, accounting for 20%. Price is the least important criterion, representing only 9% of the overall decision.
9. Identify the policy implications. In the final step, the results of the conjoint analysis are applied to the development of the implications for policy. Two levels can be considered. First, are there any lessons to be learned from the part-worth scores for an individual attribute? In the Kenyan example, as noted above, farmers exhibit a positive, though modest, preference for paraveterinarians and a higher preference for a fully qualified veterinarian (an increase in part-worth of 1.11, from 0.28 to 1.39). A comparison with the scores for the different price levels suggests that the difference in preferences between a paraveterinarian and a veterinarian could be effectively offset by a decrease in price (from high to low price) since the difference in scores between the high and low prices covers an even larger range (-2.13 to –0.71, a difference of 1.35). This suggests that, from a farmer’s perspective, paraveterinarians are acceptable if they priced right as a viable alternative to higher quality veterinary care.
Second, what can we learn from the relative importance farmers attach to the different attributes? Which attributes emerge as the most important in farmer decision-making, and are these attributes amenable to improvement? Returning to the Kenyan example, promptness of service is clearly identified as the single most important criterion, which is consistent with the highly fatal nature of the disease when the animal is not treated in a timely manner. This type of result would suggest that efforts to improve the management skills of service providers, particularly by highlighting the importance of responsiveness as an aspect of service quality, would increase the likelihood that farmers would use such services. In this case, price is found to be less important, and so adequate cost recovery should be a relatively minor issue if it is compensated by improved service quality. It needs to be remembered, though, that these conclusions relate to a very specific situation concerning curative care for a fatal disease; the conclusions will likely be different for a preventive service.
Assessing farmer preferences using the conjoint analysis technique generates useful information about the important factors – from the farmer’s point of view – that merit the attention of policy-makers and development agents. It is only one tool, though, and needs to be complemented by other approaches and techniques, such as the assessment, using contingent valuation techniques, of farmer demand and farmer willingness to pay for services. The value of such techniques is that they permit the end-user of livestock health services to reveal ex-ante the criteria and the factors that are likely to be important to them in deciding whether or not to use services and thereby avoid possible bias that may influence the factors that veterinary officers and technicians typically consider important. It may be appropriate to validate these types of studies through assessments of farmers’ actual behaviour and their use of available services.
