The Best Little Programming Tool in Extension

It seems logical that the audience for an Extension program would be identifiable and identified. How else would you discover the needs and problems for which to plan programs? How would you justify funding requests? How would you be accountable for use of funds?

Logical? Right. Nearly always the case? Wrong. Identification of specific clientele groups has become more and more difficult as audiences have increased and diversified, and more and more important as resources have diminished.

This article discusses audience identification using the small farmer as an example. Four aspects are explored: (1) how the identification process for the farmer audience has changed, (2) one way to approach the identification process, (3) how to use the information generated, and (4) implications for other programs.

A Different Era

The questions raised above about discovering farmer needs, justifying programs, and being accountable weren't generally asked until recently. The farm population was often stable over a relatively long period of time. County agents very likely knew or were known by virtually all the farmers in the county. Needs or problems were seen first-hand, as were results of change efforts. Agents could respond to farmer questions and have their performance judged as acceptable if they worked only with those who sought Extension expertise. Farmers usually formed the largest percentage of a county's population; farming was important to the economy and the Extension agent's job was seen as important to farming. Competition for resources was weak or nonexistent.

Today's agent exists in an entirely different environment. The off-farm population now outnumbers the on-farm. The relative economic importance of agriculture has declined as has Extension's unquestioned right to resources. Many agencies now compete for the same tax dollars. The farmer-to-agent ratio has increased, along with the diversification of the on-farm audience and the ensuing complexity of farmer problems.

Expectations for agents have also changed. It's no longer adequate to spend the majority of time reacting to farmer requests. Instead, agents must take a proactive stance. They must identify problems, set specific goals for change, and provide evidence of that change. Agents must justify the need for programs by clearly defining who and how many may potentially benefit from them. Today's agents need hard numbers about the entire potential clientele, not just those already using Extension's services.

Identification Process

The identification process was necessary to conduct an impact study on the two-county farming systems pilot project. The study was planned to assess the impact on the entire small-farm population in the two counties-not just on the farm families who had direct contact with program personnel. To implement this decision required a comprehensive file of small farmers in the area.

However, no such file existed at the beginning of the impact study. Sources such as the Census of Agriculture and the Agricultural Stabilization and Conservation Service (ASCS) had widely varying estimates of the total number of small farms in the area, but no one was able to provide a comprehensive listing by owner name and address. To develop such a listing, we followed six steps.

Step 1: Define Population

Any descriptors that are appropriate can be used within a given context and for which accessible information exists. A small farmer in the context of the farming systems program was defined as a person farming land within the 2-county area living outside of urban areas of 10,000 or more population and who: (1) owned 21 to 399 acres of land, (2) produced less than 15 acres of peanuts, (3) produced less than 10 acres of tobacco, and (4) raised fewer than 75,000 chickens in a single batch (fewer than 3 chicken houses). Note that each criterion is quantifiable and can be applied without bias.

Step 2: Obtain List of Land Parcels

Public tax rolls provided the list of all land parcels taxed on an agricultural exemption basis. Included were the names and mailing addresses of the owners and the size and geographical location of the parcels.

Both county tax assessor offices used the same firm for their accounting. This firm provided a separate computerized listing of the agriculturally taxed parcels to each tax assessor without any additional charge. These two lists were turned over to Extension, also at no charge.

Step 3: Eliminate Nonclientele

Agricultural exemptions are secured by many who aren't farmers. Some of these can be easily identified, for example, realty firms and timber companies. Both of these business-type parcel owners were eliminated from the farming systems study. Landowners living in large urban centers were also eliminated, because they generally lease their holdings to others. Similarly, those living outside the study area, with the exception of those in contiguous farming areas, were eliminated. Individuals living just across the county line were left on the list for they could well be farming parcels inside the county. These eliminations reduced County A's listings from 5,231 to 3,887 parcels and County B's from 4,091 to 2,106.

Step 4: Combine Parcels

The computerized listing of parcels was by parcel number, which means that any one contiguous farm unit or any one farm family's holdings could appear as many different parcels in seemingly random order on the list. Thus, it was necessary to find a way to combine holdings. This was done by entering the 5,993 remaining parcels into the files of an IBM personal computer using PeachText software.1

Groupings were made of parcels owned by the same person(s) or taxed at the same address. (Separate data entries were maintained for each co-owner of a parcel and of a farm unit.) This provided a list of 3,099 that should have been the total population of "farms" in the 2-county area. Included were landowner(s) by name(s), size of farm units, and geographical location.

Step 5: Apply Audience Definition Criteria

If criteria are applied without bias, the overall farm population can be divided into subgroups that may themselves be treated statistically as populations. To arrive at a "small farmer population," the criteria defined in Step 1 were applied. First, the computerized entries were manipulated to arrive at a listing of farm units consisting of 21 to 399 acres. Next, information secured from the ASCS office was used to eliminate farm units producing 15 or more acres of peanuts or 10 or more acres of tobacco. Similar information from the local chicken processing plant enabled the elimination of units producing more than 75,000 chickens in a single batch.

The remaining list of 2,027 farm units constituted the potential population of small farms in the 2-county area. Sixty-five percent of the potential population was located in County A and 35% in County B.

Step 6: Survey Random Sample

A final estimate of the number of "small farmers" in the potential population can be made by selecting a random sample and securing information about farming status of each unit-for example, number of acres and how acres are used.

As indicated earlier, the impetus for developing this particular small farmer file was to secure data for an impact study. The decision of the study committee was to secure data with plus or minus 10% accuracy at a 95% level of confidence. To do so required information from a probabilistic sample of 100 bona fide farm units.2

Random samples of about 100 were selected from each county's listings (a total of 196). When all names were exhausted (meaning all were accounted for in some way), questionnaires had been completed for 71 farm units meeting the "small farmer" definition. This number-71-wasn't large enough for the desired precision of plus or minus 10%, so another sample of about 100 was drawn for each county. These were exhausted in random fashion as long as resources permitted, with a target of about 65% in the sample for County A and 35% for County B. The final count was 188 attempts for A and 136 for B. After omissions were made for deceased owners, duplicate entries on the list, farm units not meeting the definition, etc., 82 small farmer units were located in County A and 45 in County B for a total of 127.

Inferences about the potential population of 2,027 could now be made between plus or minus 5.1 % and 5.5%.3 This allows for an estimate of the number of "small farmers" in the 2-county area to be between 845 and 942. Using the midpoint as the population size-894-inferences on survey questions can be made with plus or minus 8%.4

Using the Information

For the first time, Extension now knows the potential number and names of clients for the farming systems program in the two-county area. And, as a result of other question5 asked at the same time farm unit status was determined, other information is also known-for example, what they see as their problems and future plans they have for their farm acreage. Realistic projections can now be made for the number to target, what the impact should be, and afterwards, how successful the efforts have been.

Now that names and addresses are on file, periodically a small sample can be drawn, say 10 unknowns a month, and contacted, thereby adding information to the file about the overall group and tightening the precision of estimates.

The value of the list will increase in the future. As new data are added, it will be possible to discern changes that are occurring and develop a proactive stance for the coming years. Rather than react to changes that have already occurred, it will be possible to foresee trends and alter Extension's programs to prepare the farm population for the future.

Implications for Other Programs

Not every Extension audience lends itself to this type of identification process. However, more do than we generally assume. Recommendations for other program areas are:

• Criteria for selection must be quantifiable, that is, on the basis of objective data you must be able to determine if a potential client does or doesn't belong to your population.
• The criteria must make sense for programming purposes. All kinds of information are available on constituents, so make sure data selection standards are based on a sound rationale.
• Data relating to the criteria must be publicly available. You may need to be a little creative to locate sources of public information. For example, in this study, the poultry company was quite willing to provide information on their clients once they realized how it was to be used. Sources of public information that other program areas might consult include:
  • The Polk Directory - published for most metropolitan areas listing names, addresses, occupations, marital status, etc., on citizens living in the area. Public libraries or city halls usually have copies.
  • Regulatory agencies - which will often make files available. For example, the water management districts in Florida register people who use water for farm irrigation. They will provide these lists on request.
  • Schools - which maintain records on families with school-age youth. Some school districts will provide this information, while others won't. When they will, information can be secured on ages of all family members, number in family, race, parent occupation, place of residence, etc.

Conclusion

The ultimate value of developing a farmer database is that it permits the agent to more effectively meet the needs of clientele. Agents can assume an aggressive stance toward the entire clientele group rather than waiting for clientele to come to them. Such a stance provides hard data for justifying Extension's programs and decreases Extension's vulnerability. After all, Extension's best defense against attack is improved programming, which can only be done with an audience that is identifiable and identified.

Footnotes

1. PeachText 5000 (Atlanta, Georgia: Management Science America Inc., 1985).

2. S. Sudman, Applied Sampling (New York: Academic Press, 1976).

3. The original sample equals 324: deceased and duplicate units equal 17, nonrespondents equal 38, respondents equal 269 (127 farmers and 142 nonfarmers). A liberal estimate of the percentage of "small farmers" among the 2,027 was achieved by dividing the number of small farmers located (127) by the number of respondents (269) or 127 = 269 = 47.2%. A conservative estimate was attained by dividing the number found by the total sample less the deceased and duplicate units or 127 = 307 = 41.4%. Using Formula A and Formula B from M. F. Smith, Sampling Considerations in Evaluating Cooperative Extension Programs, PE-1 (Gainesville, Florida: University of Florida, Cooperative Extension Service, IFAS, 1983), these 2 percentages were estimated to be accurate at ± 5.5% and ± 5.1 %, respectively, with 95% confidence. The overlap is the area on a continuum that both estimates cover, which equals 41.7°/a to 46.5% x 2,027 = 845 to 942, arrived at as follows:

41.4 ± 5.1% = 36.3% to 46.5%
47.2 ± 5.5% = 41.7% to 52.7%
Overlap = 41.7% to 46.5%
X 2027 = 845 to 942
Midpoint = 894.

4. Using the same formulae as in footnote 3, the sample of 127 can be used to draw inferences about the population of 894 small farmers with an estimate precision of ± 8% at 95% confidence.

5. After it was determined that a landowner was a "small farmer," other questions were asked to meet impact study requirements.