Binary Logit Model Results 50

The concordant figure is 74.4 percent for the binary logit model. Bounded between zero percent and 100 percent, the concordant is parallel to an R2 value in a linear model. The estimated model compares pairs of predicted values with the actual observation so that if the observation at step j – k has a lower predicted value than the predicted value at step j the pairs are said to be

“concordant” or consistent with each other. In the case of the binary logit model, there are 621 observations, so 621*(620)/2 = 192,510 possible pairings. The model correctly predicted when step j-k was actually less than step j for 74.4 percent of the pairings which is considered a good fit for cross-sectional data.

Column two shows the parameter estimates while column three has the standard errors in table 6.1. Note that the logit model has a single intercept. The dependent variable measures whether the VAPG recipient reached the first steps 1 through 8 in business development process.

51

The coefficient on the intercept is significant indicating that there is unique information contained in the first eight steps and the last step of business development. Other significant coefficients are on the variables BC, GRANDPRO, SALENPRO, and MARKETSHAR. Significance of parameter estimates were observed for seven of the nineteen crop variables (DAIRY, FLOW, FRUIT, NUTS, SMEAT, WHEAT, and WINE) and one of the four business organizational forms (APGROUP). Finally, significant parameter estimates were found for the binary state variables of Illinois (IL), Kansas (KS), Minnesota (MN), Missouri (MO), and Wisconsin (WI).

BC denotes the number of USDA Rural Business and Cooperatives division employees in the state where the VAPG recipient resides and is a measure of resources available to assist the VAPG recipients. The negative sign suggests that as the number of employees increases, the likelihood of observing a VAPG recipient in the first eight steps decreases. Correspondingly, the likelihood increases for observing a VAPG recipient in the last step of business development with a successful product being marketed in March of 2006. It is not possible to obtain precise information on the number of employees in each state over time and their individual job responsibilities. However, anecdotal information collected by the principal investigator on the grant suggests that USDA Rural Development increased the number of Rural Business and Cooperatives division employees and refocused job responsibilities in order to help manage and work with the VAPG program after its authorization in the Farm Bill. This would suggest that the quantity of a marketing input, marketing expertise, increased for these VAPG recipients which would lead to a decrease in the marketing margin and thus, accomplishing one of Congress’s goals for this program.

52

GRANDPRO denotes the VAPG grant amount measured in dollars divided by the

number of producers that own the organization that received the VAPG grant. SALENPRO is the sales volume for the organization that received the VAPG grant divided by the number of

producers. These variables are a measure of size and their coefficients had negative signs. An increase in the value of grant dollars received or sales volume for the VAPG recipient in the numerator (or a decrease in the number of producers in the organization in the denominator) suggests that the likelihood of observing a VAPG recipient in steps one to eight decreased. Alternatively, the likelihood increases for observing the VAPG recipient in the last step of business development.

It is difficult to make any broad generalizations regarding these variables. However, larger VAPG grants tended to go to organizations that had a successful business operation with existing sales volume and were seeking to expand into a value-added product which would suggest that such firms had good intelligence regarding the market for such a product. Very few large grants went to businesses that were starting a value-added product from “scratch.” This observation would suggest that these firms knew that the demand for the value-added product was increasing which would lead to a decrease in the marketing margin which was a goal of the VAPG program.

MARKETSHAR measured the proportion of the commodity produced in the county where the VAPG recipient was located divided by the total U.S. production of that commodity. This variable is a measure of the underlying commodity being utilized and its coefficient had a negative sign, which suggests that as the market share increased (through an increase in the numerator which would suggest greater production in that local market or a decrease in the denominator which would suggest a smaller national market), the likelihood that the VAPG

53

recipient was in one of the first eight steps decreased. The same intuition for the previous two variables exists with this variable in that the likelihood that the VAPG recipient was successful in marketing the product in March of 2006 was observed. Thus, the marketing margin decreases when the supply of the commodity input increases, which was a goal of the policymakers for the VAPG program.

Crop binary variables that had significant coefficients included DAIRY, FLOW, FRUIT, NUTS, SMEAT (specialty meats), WHEAT, and WINE. The parameter estimates were negative for these variables suggesting that the VAPG recipients adding value to these commodities relative to OTHER (which was the dropped binary variable) had a decreased likelihood of being in steps one to eight, or rather an increase in the likelihood that these VAPG recipients were in step nine with a product being marketed in March 2006. It should be noted that the coefficients on PORK and SGRAIN (small grains such as mustard, buckwheat, and other grains) were significant at the 0.20 level of significance with a negative sign.

One of the four business organizational forms (APGROUP) was significant with a

positive sign which would suggest that a successful VAPG grant written by this organization had an increased likelihood of being in the business development steps of one through eight relative to MAJCON which was the dropped binary variable. Remember that APGROUPs are trade associations composed of producers or cooperatives. These organizations tend to have a membership that is very broad and diverse. Furthermore, these variables do not undertake business development but rather make the results of their VAPG grant available to all their members to consider developing a business for the opportunity identified by the study. Many of these activities are market studies. Thus, this result may not be that surprising. It should be noted that the number of VAPG grants awarded to APGROUP declined in every year from 2002 (91

54

grants) to 36 in 2005 which would suggest that these entities were not as successful in receiving VAPG grants or that they did not submit as many grant proposals in later years.

Significant and negative parameter estimates were found for the binary state variables of Illinois (IL), Kansas (KS), Minnesota (MN), Missouri (MO), and Wisconsin (WI). These results suggest that VAPG recipients located in these states had a decreased likelihood of being in steps one to eight or alternatively, VAPG recipients in these states had a greater likelihood of

marketing a product by March 2006. Illinois had twelve (18 total grants or $2,558,000 in step nine) grant recipients reach step nine with an average VAPG grant value of $213,167. Kansas had ten (18 total grants or $2,738,340 in step nine) grant recipients that reached step nine, with an average grant value of $273,864. Minnesota had eighteen (20 total grants or $3,655,930 in step nine) grant recipients that reached step nine, with an average grant value of $203,107. Missouri had twenty four (45 total grants or $5,192,220 in step nine) grant recipients that reached step nine, with an average VAPG grant value of $216,344. Wisconsin had eighteen (22 total grants or $2,515,554 in step nine) grant recipients reach step nine, with an average grant value of $139,753.

It is hard to know what these results suggest. The principal investigator on this grant that obtained the information on the dependent variable noted that Missouri probably has the most sophisticated infrastructure for business development with a long-standing state program and each recipient is “strongly encouraged” to receive education after receiving a state VAPG grant or a USDA VAPG grant. Missouri is the only state to have such a “requirement.” It must also be noted that the states are all located in the Midwest and are contiguous to one another. Illinois, Missouri, and Wisconsin have a state VAPG program. Further research should be done to

55

discern why VAPG recipients in these states were more successful in having a product marketed in March 2006 relative to other states.