volume 57 article #6

Factors Affecting Commercial Bank Use of FSA Loan Guarantees in Arkansas

Bruce L. Dixon, David L. Neff, Bruce L. Ahrendsen, and Scott M. McCollum

Dixon, Neff, and Ahrendsen are professor, associate professor, and associate professor of agricultural economics, respectively, and principals of the Center for Farm and Rural Business Finance, and McCollum is a former graduate assistant, all with the University of Arkansas at Fayetteville. McCollum is a recipient of the Farm Credit Foundation Scholarship Program, which partially supported his graduate research. This study was partially supported by USDA-CSREES Agreement No. 95-34275-1319. The authors wish to thank Gary Groce of the state FSA office in Little Rock, Arkansas, and Steve Ford of FSA in Washington, DC, for providing the data and background information. The authors claim sole responsibility for any remaining errors and for the views expressed here.

Abstract

The Farm Service Agency (FSA) guaranteed loan program has existed for more than two decades, and it will continue under the FAIR Act to be a major support program for farm operators. This study identifies factors influencing Arkansas commercial bank participation in FSA loan guarantee programs and the dollar volume of participation during fiscal years 1990­95. Incidental truncation models are estimated to predict whether commercial banks participate in making guaranteed operating loans or farm ownership loans, as well as to predict obligation volumes. Agricultural banks are found to be larger users of loan guarantees than urban and nonagricultural banks. Increasing loan-to-asset ratios are associated with greater bank participation. Belonging to a multi-bank holding company encourages use of farm ownership loan guarantees.

Key words: FSA guaranteed loan program, commercial banks, loan obligation levels, Farm Service Agency.

Article <top>

United States agricultural policy is entering a new era because of the passage of the Federal Agricultural Improvement and Reform (FAIR) Act of 1996. A highly important component of the FAIR Act is the elimination of the linked price support programs for many commodities. The Farm Service Agency (FSA) guaranteed farm loan program is one of the major remaining financial programs assisting farmers. Under the FAIR provisions, authorized guaranteed loan levels are set to rise from $2.5 billion in fiscal 1996 to $2.85 billion in fiscal 2000 (Koenig). Over this period, direct loans are constrained to a maximum of $.585 billion annually.

Historically, the FSA (formerly the Farmers Home Administration) was very active in direct lending. In 1984, FSA lending policy was switched from emphasizing direct lending to guaranteeing loans originated by other financial institutions as a means of lowering direct costs to the federal government, processing loans more efficiently, and lowering loan losses (Koenig and Sullivan). Loan guarantees increased from 2% of total FSA obligations in fiscal 1983 to 52% in fiscal 1987 (Ahrendsen). In fiscal 1996, the guaranteed portion increased to 69% [U.S. Department of Agriculture/Economic Research Service (USDA/ERS)]. Direct loans, although still an important component under the FAIR Act, receive lower apportionments for obligations and have more limitations on borrowers than previously (Koenig).

Guaranteed loans are made to borrowers unable to obtain credit without government assistance. Financial institutions—primarily commercial banks and the Farm Credit System (FCS), but also other institutions such as co-ops and mortgage companies—can obtain guarantees for up to 90% of the loan value.¹ Farm operating loans (OL) for purposes such as annual operating costs, equipment purchases, or refinancing can receive guarantees for up to $400,000 of principal. Farm ownership (FO) loans can obtain guarantees for up to $300,000 of principal for land purchase and improvement. Under the FAIR Act, many provisions were included to increase use by beginning farmers (Koenig).

¹The maximum guarantee was increased to 95% by the FAIR Act for certain loan purposes, such as for beginning farmers or for refinancing of FSA direct loans by private loans (Koenig).

Given the continued importance of guaranteed loans outlined in the FAIR Act, it is important to ascertain the factors which motivate financial institutions to use guaranteed loans and the level of that use. In this study, data from Arkansas are used to identify variables that affect a bank’s decision to use the guaranteed loan program for operating and/or farm ownership loans. These decisions are modeled in the framework of an incidental truncation model (Greene).

The purpose of this study is to expand the understanding of how changes in bank structure, such as merger activity, lessening of the number of agricultural banks, and changing economic conditions, might affect use of guaranteed loans. Under the FAIR Act, the riskiness of returns to agriculture may increase. This could cause more fluctuation in incomes and collateral values, increasing the demand for guaranteed loans by lending institutions. As shown in this study (and supported by a national study by Koenig and Sullivan for fiscal year 1988), a substantial proportion of institutions eligible to use guaranteed loans do not, so there is the possibility of expanded bank participation. The estimated models could be used to predict such effects.

FSA Guaranteed Loan Activity in Arkansas <top>

An understanding of the banking environment in Arkansas is important for the analysis in this study. During the model estimation sample period (fiscal years 1990­95), Arkansas had unit banking with branching within a county. On January 1, 1994, branching into contiguous counties was allowed. Usury laws require that the interest rate on loans by commercial banks not exceed the federal discount rate by more than 500 basis points. During the sample period, Arkansas had approximately 258 banks; of these, about 40% satisfy the Federal Reserve System’s (FRS’s) definition of agricultural banks.²

^{2 The FRS Board of Governors classifies a bank as agricultural "if its ratio of farm loans to total loans exceeds the unweighted average of the ratio at all banks" (USDA/ERS, p. 10) as of a certain date. The December 31 cut-offs were used in this study.}

The number and dollar volume of loans guaranteed in Arkansas by the FSA clearly demonstrate that the program was not extensively utilized in that state until the mid-1980s (as discussed in McCollum). Fiscal year 1985 marked the beginning of greater national funding and comparatively larger numbers of guarantees in Arkansas with 93 guarantees compared to 16 in 1984. As shown in Table 1, the level of FO obligations³ in Arkansas is roughly a half to a third of OL obligations in a given year, with FO loans showing growth over the fiscal 1990­95 six-year period. In contrast, OL obligations have not shown such growth. On a national level, FO obligations have been about a third of OL obligations, so that Arkansas has a slightly higher proportion of FO to OL obligations than the U.S. Arkansas has mirrored the U.S. growth in FO obligations, but has not shown a substantial growth in OL loans compared to the U.S. In terms of percentages of delinquent loans, Arkansas is clearly lower than the U.S. for FO loans in general, and also lower than the U.S. for OL loans in fiscal 1993­95.

³ Obligations are the amounts of principal of the guaranteed loan. The percentage of the loan guaranteed can vary.

 

The ratio of OL loans to FO loans in Arkansas over the sample period was roughly three to one. For the U.S., the ratio lies between four to one, and five to one over the years 1991­95.⁴ During 1990­95, the annual average OL obligation in Arkansas ranged from $114,000 to $134,000, and the FO annual mean obligations ranged from $171,000 to $208,000. Arkansas OL levels were somewhat higher than national averages; U.S. figures for 1991­95 ranged from $100,000 to $106,000. National FO loans averaged $151,000 to $160,000 during this period, so the differences between Arkansas OL and FO means were somewhat larger than for the U.S. as a whole. Commercial banks were the predominant originators of both OL and FO loans in Arkansas, just as with the nation as a whole.

Types of agricultural enterprises vary strikingly across Arkansas, with the east growing field crops and the west producing poultry and cattle. Over 83% of the 1990­95⁴ OL loans were made in the three easternmost crop-reporting districts in Arkansas. No such state regional dominance occurred with the FO obligations, although there is considerable disparity among frequencies across crop-reporting districts.

⁴The data in this paragraph were provided by Steve Ford of the FSA Washington, DC office.

In summary, Arkansas has been active in using the guaranteed loan program, and its levels of obligations and delinquency rates are not out of line with the rest of the nation. Arkansas has a diverse agriculture with concentrations in both a variety of field crops and livestock enterprises. Thus the findings of this study likely provide useful insights for the guaranteed loan program throughout the U.S.

General Modeling and Estimation Approach <top>

Two, two-equation submodels are hypothesized. One submodel describes bank participation in OL loans, and the other describes bank participation in FO loans. Each submodel consists of a probit (selection) equation and a regression equation of the form:

(1)

(2)

where z_it is an indicator (0­1) variable denoting whether or not bank i made a loan in year t, and y_it is a continuous variable representing obligation volume by bank i in year t. The vectors X_it and W_it represent independent variables, and the vectors β and γ contain parameters to be estimated. The error terms ε_it and η_it are normally distributed with means zero and variances of unity and σ², respectively, and they may be jointly distributed. Equations (1) and (2) comprise an incidental truncation model as discussed in Greene.

The two submodels pertain to OL and FO loans, respectively. In each submodel, the first equation (selection equation) portrays the decision of a bank to use loan guarantees in a given year. The second equation in each submodel explains variation in the level of obligations incurred. This is the "regression equation" in the incidental truncation model.

The selection equation is estimated as a probit model since the dependent variable is binary in both submodels. The main distinguishing feature of incidental truncation models is that observations on the dependent variable in the regression equation are available only if a bank decides to use loan guarantees. This may result in incidental truncation. Hence, to obtain consistent estimates in the regression equation, a two-step approach is used (as in Dixon, Ahrendsen, and Barry; or Greene).

Initially, the same variables are hypothesized to occur in both equations of a given submodel because those variables motivating use of the guarantees are also potentially important in determining the levels of guarantees and vice versa. However, in the estimation phases of this study, only those variables showing some significance in a particular equation (discussed below) are retained in the final version of that equation.

Factors Affecting Use and Levels of Loan Guarantees <top>

Koenig and Sullivan investigated the characteristics of banks making agricultural loans in 1988, and separated them into one group not using any guarantees and two other groups that did use guarantees. Their analysis suggests that the agricultural loan/total loan ratio (AGTL), return on assets (ROA), and asset levels (ASSET) are positively related to a bank utilizing guaranteed loans.

The lender’s propensity to invest available funds in loans, as opposed to other investments, is measured by the loan-to-asset ratio (LAR) of a bank. An aggressive lending policy increases LAR, while simultaneously expanding the bank’s exposure to loan losses. It is hypothesized that LAR is positively related to the bank’s volume of FSA guaranteed loans.

A significant variable in predicting variation in bank agricultural loan market share is whether the bank is in a metropolitan statistical area (MSA) as defined by the U.S. Office of Management and Budget (Ahrendsen, Dixon, and Priyanti). Eleven of the 75 counties in Arkansas are currently in MSAs. MSA is a binary variable taking on the value of one if a bank is located in an MSA. MSA is hypothesized to have a negative relationship with the number and volume of FSA guaranteed obligations since rural counties are usually more reliant on farm income than their urban counterparts.

A bank’s competitive position vis-à-vis other banks in a given market may affect how actively a bank seeks out borrowers, and how aggressive a bank’s lending activities must be in order to increase loan volume for a given demand (Sullivan). The level of competition is reflected by market share (MS), which is calculated as the proportion of total bank deposits held by a bank in its county, or MSA if the county is in an MSA. A bank usually confines its activities to a 25­30 mile radius of its office, so that it experiences its greatest competition from banks in close proximity (Rose). The Herfindahl-Herschmann Index (HHI) also is used as a means of measuring competition in the bank’s market (county unless the bank is in an MSA, then MSA is the market). The HHI measures the concentration of deposits of the banks in a market. HHI increases as concentration of deposits among fewer banks in a market increases. HHI and MS are expected to be negatively related to a bank’s volume of FSA guaranteed loans because the loans allow banks in competitive markets (small HHI, low MS) to compete more aggressively by making loans to marginal borrowers. However, a positive relationship could occur due to loan risk being concentrated among fewer banks.

As a bank experiences relatively heavy losses on a particular investment type compared with the rest of its asset portfolio, it likely would evaluate the efficacy of continuing to allocate available funds to that investment and seek ways to decrease the risk of such investments. RISK is the ratio of total net agricultural loan losses to total outstanding agricultural loans divided by the ratio of total net loan losses to total outstanding loans for a bank. The loan losses are net of any loan payments made after the loan has been previously written off a bank’s loan portfolio. The variable RISK represents the relative riskiness of agricultural loans to nonagricultural loans. The a priori relationship of RISK to the volume of FSA guaranteed loans is ambiguous. Higher levels of RISK could cause curtailment of agricultural loans if banks want to limit their exposure to risk from agriculture, implying the sign is negative. But if rising RISK encourages more use of guaranteed loans due to banks’ desire to continue serving agriculture in risky times, the sign is positive.

The affiliation of a bank with a multi-bank holding company (MBHC) means direct access to a correspondent bank(s). This allows banks to diversify the risk of a given loan by having a correspondent bank participate in a loan at generally lower transactions costs than if the correspondent bank were not a member of the bank holding company. A value of MBHC equal to one implies a bank is a member of a multi-bank holding company (zero otherwise), and the direct-access-to-correspondents effect implies a negative relationship to the volume of FSA guaranteed loans. Conversely, an MBHC may have the resources to offer its banks specialized processing of guaranteed loans, so MBHC would be positively related to FSA guaranteed loan use. Thus the relationship of MBHC to FSA guaranteed loan use is ambiguous.

A measure of collateral risk is the coefficient of variation of the value of farmland and buildings in the county where the bank is located (CVFARMV) based on the four previous years. CVFARMV is hypothesized to be positively related to FSA guaranteed obligation volume. DeVuyst, DeVuyst, and Baker hypothesized that volatility in land values requires a risk premium to be paid by the borrower. The loan guarantee can limit the risk of a marginal borrower without requiring a risk premium as compensation for the lender. Most likely, CVFARMV would be more important to FO loans since they are more typically collateralized by land. However, declining long-term asset values can reflect general financial hardship in agriculture and the difficulty of paying back all types of loans.

The proportionate change in farm income from one year to the next year by county (ΔFMINC) was found by Ahrendsen, Dixon, and Priyanti to be positively related to bank market share of agricultural loans. Rising farm income may imply more loanable funds and an inclination of banks to lend to agriculture. Also, increased farm income causes demand for loans to increase since farmers are better able to qualify for higher loan amounts and may wish to expand their enterprises. However, increased income also allows for self-financing, so ΔFMINC has an ambiguous sign expectation. In the present study, ΔFMINC is defined for the county where the bank is located. In addition, greater variability in farm income increases the risk associated with lending to the agricultural sector. Therefore, CVFMINC, the coefficient of variation over the previous four years in net farm income per county where the bank is located, is expected to be positively related to volume of FSA guaranteed obligations.

Different types of agricultural enterprises (crops versus livestock) traditionally have had different credit demands. To reflect this diversity, the ratio of revenues from the sales of field crops to total agricultural revenues in the county of the bank’s location (FCREV) is defined. It is expected that FCREV is positively related to the volume of OL guaranteed loans since field crop farms typically have more demand for operating capital than do poultry and beef cow-calf farms. It is not clear what the sign should be for FO loans.

An approved (ALP) or certified (CLP) lender program designation by the FSA means that the bank is an active agricultural lender. The bank must have qualified personnel, an acceptable loss rate and/or have originated at least a minimum amount of guaranteed loans. In return, the bank incurs lower transactions costs. Such banks have an incentive to use FSA loan guarantees in order to retain their ALP or CLP status. We define a binary variable to represent these preferred lenders (PREF) to have a value of one if the bank has ALP or CLP status, and zero otherwise. The coefficient is expected to be positive.

As the real interest rate charged on loans increases, it is more difficult for borrowers to qualify for credit given their existing payment capacity. Increased loan payments can lead to financial failure, as noted by Shephard and Collins. To offset this risk, the lender could obtain a guarantee on the loan to lower asset risk. Therefore, INT, the discount rate plus 475 basis points, adjusted for inflation, is expected to be positively related to the frequency and volume of FSA guaranteed loans.⁵

⁵ The Arkansas usury law stipulates that the interest rate on loans can be no more than the federal discount rate plus 500 basis points. The interest rate in Arkansas is typically near this limit, although the limit varied over the fiscal years 1990­95.

Variable Construction and Data Sources <top>

Data for variables used in the analysis are annual observations. All variables except MS and HHI are for an entire year. MS and HHI are calculated based on June levels of deposits each year. All the remaining variables except for CVFARMV, CVFMINC, ΔFMINC, and FCREV are computed using fiscal year observations. Because CVFARMV, CVFMINC, ΔFMINC, and FCREV are observed over calendar years, they are all lagged one year in the models. RISK is lagged one year to allow adjustment time for changing risk conditions.

Two binary dependent variables are needed for the probit models to indicate if a bank made FO or OL loans in a given year. Thus the dependent variable for the OL probit model, OBL, equals one if a bank made an OL loan in a given fiscal year, and zero otherwise; the dependent variable for the FO probit model, OBF, is defined similarly for FO loans. The two continuous dependent variables for the regression models, OLOBL and FOOBL, represent total OL and FO obligations, respectively, in a given year by a bank in thousands of dollars.

The data are drawn from several sources: the FSA state office in Little Rock, Arkansas; Federal Deposit Insurance Corporation (FDIC) quarterly call reports of income and condition, and summary of deposits; U.S. Department of Commerce, Bureau of Economic Analysis (BEA); and the Bureau of the Census.

The data are a time series of cross-sections. The time period begins in fiscal 1990 and ends in fiscal 1995. All financial data are deflated by the CPI-U (1982­84 = 100). Because complete data were not available for all banks in the state, the sample had to be reduced to 243 banks (although observations were not always available for a given bank in all six years). Thus the sample for the two submodels (OL and FO obligations) consists of 1,423 observations over the six-year period.

Estimation Procedures <top>

The number of hypothesized independent variables in each of the two submodels is quite large. Two specification approaches are reasonable in such a situation. One approach is to estimate the models as specified and present the results. Such an approach has the advantage of minimizing any pre-test bias. A deficiency of this approach is that some efficiency of estimation is lost because many coefficients whose true values are zero are left unrestricted. A second approach is to start with a large number of regressors and reduce it to a more specific model on the basis of statistical testing. This general-to-specific approach is discussed in Maddala.

To capture the efficiency of restricting irrelevant regressors to have zero coefficients, the general-to-more-specific model approach was adopted. First, the models were estimated with all hypothesized variables included. Next, all explanatory variables in the probit and regression equations that had a calculated absolute value of the ratio of the estimated coefficient to its estimated asymptotic standard error (z) of less than one were eliminated and the models were reestimated. The criterion of dropping a variable if its z-value is less than one in absolute value is similar to the model specification rule for maximizing adjusted R ². The application of this rule to probit models has not been established in the literature as a specification procedure, but we assume it to be a good compromise between deleting all variables not significant at the .05 or .01 level, or including all regressors regardless of significance.

A final estimation consideration is heteroskedasticity in the probit equations which, if it exists, can result in inconsistent estimates. Unfortunately, there is not a broad set of tests in the literature for the existence of heteroskedasticity in probit models. As Greene points out, the discovery of heteroskedasticity in the probit model may indicate an omitted variable. Because of this potential, ASSET was left in the two loan volume probit equations to guard against a possible misspecification, and because assets seemed the most likely source of heteroskedasticity in the probit models.

Since the data are panel in nature, the possibility arises that the error terms of the regression equations in the incidentally truncated models were heteroskedastic. In fact, even if the regression equation without the inclusion of the inverse Mill’s ratio (IMR) is homoskedastic, inclusion of the IMR induces heteroskedasticity (Greene). Because of this, the covariance matrices of the regression equations’ parameters were estimated using White’s heteroskedasticity-consistent covariance matrix. Thus, the estimation procedure was to estimate the selection equation by probit and use the parameters of the estimated probit equation to estimate the IMR. Then the IMR was included as a regressor in the regression equation and the regression coefficients were estimated by OLS and their standard errors computed using White’s formulae.

 

Descriptive Statistics of Sample Banks <top>

Table 2 provides a brief summary of similarities and differences between banks that made no FSA OL or FO loans in the sample and those that did. In agreement with Koenig and Sullivan, we found that higher levels of ASSET, AGTL, and MBHC are associated with participating banks, although the margins for ASSET and MBHC are not large. Banks in counties with more of their agricultural revenues from field crops are more likely to have participated. Loan-to-asset ratios and ROA are not important distinctions between participants and nonparticipants. Fewer than half of the 243 banks in the estimation sample (only 108) made at least one guaranteed loan during fiscal years 1990­95.

Among the 108 banks that made one or more guaranteed loans over the six years, 67 had a mean AGTL equal to or greater than 17%.⁶ Of the 101 banks with mean AGTL greater than or equal to 17%, 34 of them did not make a single guaranteed loan. Thus agricultural banks were more likely to use guaranteed loans than nonagricultural banks, but not all agricultural banks used guarantees and 29% of nonagricultural banks did make at least one guaranteed loan.

⁶ Seventeen percent is the 1996 cut-off between agricultural and nonagricultural banks used by the FRS Board of Governors (USDA/ERS).

Estimated OL and FO Submodels <top>

Initially, the two obligation submodels hypothesized had 15 variables plus the IMR in the regression equations. In the first estimation, with all variables included, the OL probit model had three variables with z’s less than one in absolute value; the OL regression equation had five such variables.⁷ In the FO probit model, six variables were removed in the first round estimation and 12 variables were removed from the regression equation.⁸

⁷ For the OL probit model, the variables ΔFMINC, CVFMINC, and MSA were eliminated. In the OL regression model, the variables RISK, ΔFMINC, CVFMINC, PREF, and the inverse Mill’s ratio were eliminated.

⁸ In the FO probit model, RISK, HHI, ΔFMINC, CVFMINC, ROA, and MSA were eliminated. For the FO regression model, the variables ASSET, LAR, AGTL, MS, FCREV, ΔFMINC, CVFARMV, CVFMINC, INT, ROA, PREF, and the inverse Mill’s ratio were eliminated.

The second round estimation generated the final models. The overall estimation results for the four equations in the obligation submodels display some explanatory power and statistical significance. Probit model classification power is good in the sense of correctly classifying 83% and 90% of the OL and FO sample observations, respectively. These high rates are not surprising because most banks did not use loan guarantees in a given year. However, both models classified better than if the impact of independent variables were ignored. Both probit equations have significant variables, indicating explanatory power. The two regression equations vary considerably in terms of R². The OL volume equation has an R² of 0.186, somewhat low for primarily cross-sectional data, but the FO volume equation has an R² of 0.370, respectable for cross-sectional data.

Elasticities of variables significant at the .05 level on a two-tailed test are presented in Table 3 for the continuous variables, and the coefficients for the binary variables that are similarly significant are presented in Table 4.⁹ Elasticities are used because one dependent variable in each submodel is qualitative (probit) and one is quantitative (regression). Elasticities are not relevant for binary variables, and so their actual coefficients are presented. Note that the binary variable coefficients in the probit models are the number of standard deviations the mean of the probit function increases when the binary goes from zero to one. Thus, such coefficients can be compared with each other, and those larger in absolute value have more impact on the probability of using loan guarantees.

⁹ These are direct elasticities. Total regression equation elasticities of variables that appear in both selection and regression equations in a given submodel include the effect of changes in the IMR (see Greene). However, the IMR was eliminated in the first round of the estimation of the OL and FO obligation models.

The probit models have the larger numbers of significant variables. The coefficients of LAR, AGTL, HHI, and PREF have the anticipated signs in the OL probit model. The signs on LAR, AGTL, and HHI indicate loan guarantees are viewed as a risk-reducing tactic. The positive sign on PREF shows that banks with this designation are more likely to use the loan programs. The negative sign on HHI indicates that as deposit concentration in a bank’s county or MSA increases, the bank is less likely to use loan guarantees. However, as a bank’s individual share of the county or MSA’s deposits increases (MS), it will be more inclined to use guarantees—contrary to our original hypothesis. Thus there are countervailing competitive forces at work. Increased deposit concentration lowers demand in a county or MSA for OL loan guarantees, but individual banks, ceteris paribus, will utilize guarantees as their market share of deposits increases. The positive sign on FCREV reflects the fact that counties whose primary agriculture is crops had by far the largest share of OL guarantees, both in numbers and dollar volume.

 

Many of the variables that are significant in the OL probit model are also significant for the FO probit model. Only MS has an unexpected sign, similar to the OL probit model. Bank size (ASSET) is significant but highly inelastic. The other significant continuous variables, LAR, AGTL, and INT, indicate loan guarantees are viewed as a risk-reducing activity. These three variables have elasticities of 0.90 or higher. Not surprisingly, FCREV is negative, reflecting the fact that FO loans are frequently used in areas not dominated by field crop agriculture. Membership in a bank holding company encourages the use of loan guarantees. This favors the hypothesis that a multi-bank holding company offers its members specialized processing of loan guarantees. Not unexpectedly, preferred lender status indicates an increased likelihood of utilizing FO loan guarantees.

In the FO volume equation, the RISK elasticity is positive and the HHI elasticity is negative, the latter consistent with its sign in the OL probit submodel. The positive sign on RISK suggests that as agricultural loans become more risky, banks want to limit exposure of potential loan losses by having guarantees for longer term loans. The most elastic variable is HHI, which indicates that as markets become more competitive (HHI decreases), the volume of guarantees increases. Once again, rural banks are more likely to have a higher volume of guarantees, and membership in a multi-bank holding company will increase volume of guaranteed obligations.

Summary and Conclusions <top>

Over the last decade, the guaranteed loan program has largely replaced the direct loan program at FSA as a means of providing financing for production agriculture. Under the provisions of the FAIR Act, guaranteed loans will be a more important source of government support for production agriculture, with their levels of obligations set to increase from current levels. This study identifies those factors that determine which banks utilize operating and farm ownership loan guarantees in Arkansas.

Guaranteed loans are clearly a risk-reduction tool for Arkansas commercial banks. The estimated models indicate that, ceteris paribus, banks with relatively larger proportions of agricultural loans were more likely to use loan guarantees and, in the case of OL loans, have larger guarantee volume. Increases in real interest rates encouraged the likelihood of a bank using FO loan guarantees.

Bank size, although statistically significant, did not substantially impact loan guarantee use, and so the economies-of-size hypothesis for using guaranteed loans is not strongly supported. However, economies of size may be more apparent via the multi-bank holding company effect of being able to provide expertise in using guarantees. Also, approved or certified lender status is associated with a higher propensity to use loan guarantees, as expected. Rural banks were more inclined to use larger levels of guarantees, if they used them, than were urban banks. Interestingly, economic conditions, such as the change in net farm income and variability of net farm income and land values, were weak explanatory variables. Thus, if the impact of the FAIR Act is to increase financial volatility in production agriculture, it might not result in increased use of guaranteed loans. However, since the first half of the 1990s were relatively stable, this conjecture needs to be tested over more volatile eras.

The FSA guaranteed loan program appears to be attractive to rural, agricultural banks in Arkansas, particularly in field crop-intensive counties for OL loans. Membership in a multi-bank holding company resulted in increased use of guaranteed loans; thus bank consolidation does not imply that the program will become obsolete as banks continue to merge. What appears to be crucial is the continued existence of banks specializing in agricultural lending, since agricultural banks are more likely to use loan guarantees than nonagricultural banks.

If changes in bank structure lead to fewer banks specializing in agricultural lending, fewer banks may use the guaranteed loan program. If these remaining banks became even more concentrated in agriculture, they might make a higher volume of guaranteed loans than is presently the case. They would likely avail themselves of preferred lender status which would allow guaranteed loans to be processed more efficiently. Also, Farm Credit System institutions might become larger participants if there were fewer agricultural banks. Further research relating Farm Credit System participation in guaranteed loans to the proportion of agricultural banks would yield more solid evidence on this possibility.

Arkansas reflects the diversity in U.S. agriculture as a whole because of its balance between both animal and crop agriculture. Its overall participation patterns in the guaranteed loan program are similar to those of the U.S. as a whole. Hence, the results of this study are suggestive for bank behavior in the U.S., but a broader study covering multiple states would be useful to confirm the present findings.

References <top>

Ahrendsen, B.L. "Federal Legislation: Agricultural Credit Issues and Institutions." Spec. Rep. No. 171, Arkansas Agr. Exp. Sta., Fayetteville, November 1995.

Ahrendsen, B.L., B.L. Dixon, and A. Priyanti. "Growth in Agricultural Loan Market Share for Arkansas Commercial Banks." J. Agr. and Appl. Econ. 26 (1994):430­42.

DeVuyst, L., E. DeVuyst, and T. Baker. "Expected Farm Mortgage Default Cost." Agr. Fin. Rev. 55(1995):10­22.

Dixon, B.L., B.L. Ahrendsen, and P.J. Barry. "Explaining Loan Pricing Differences Across Banks: Use of Incidentally Truncated Regression." Agr. Fin. Rev. 53(1993):15­27.

Greene, W.H. Econometric Analysis, 2nd ed. New York: Macmillan Publishing Co., 1993.

Koenig, S.R. "Credit." In Provisions of the Federal Agricultural Improvement and Reform Act of 1996, edited by J. Nelson and L.P. Schertz. Agr. Information Bull. No. 729. USDA/ERS/CAD, Washington, DC, September 1996.

Koenig, S.R., and P.J. Sullivan. "Profile of Participants in FmHA’s Guaranteed Farm Loan Programs." Staff Rep. No. 9160. USDA/ERS/Agriculture and Rural Economy Div., Washington, DC, December 1991.

Maddala, G.S. Introduction to Econometrics, 2nd ed. New York: Macmillan Publishing Co., 1992.

McCollum, S.M. "Determinants of FSA Guaranteed Loan Use Volume and Loss Claims for Arkansas Commercial Banks." Unpublished M.S. thesis. University of Arkansas, Fayetteville, 1996.

Rose, P.S. Commercial Bank Management, 2nd ed. New York: Irwin Publishing Co., 1993.

Shephard, L.E., and R.A. Collins. "Why Do Farms Fail? Farm Bankruptcies, 1910­78." Amer. J. Agr. Econ. 64 (1982):609­15.

Sullivan, P.J. "The Structure of Bank Markets and the Cost of Borrowing: Evidence from FmHA Guaranteed Loans." In Regulatory, Efficiency, and Management Issues Affecting Rural Financial Markets, edited by C.-B. Moss. Pub. No. SP93-22. Food and Resour. Econ. Dept., University of Florida, Gainesville, September 1993.

U.S. Department of Agriculture, Economic Research Service. Agricultural Income and Finance: Situation and Outlook Report. Pub. No. AIS-64. USDA/ERS, Washington, DC, February 1997.

White, H. "A Heteroscedastic-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity." Econometrica 48(1980):817­38.

<top>

 

---

Send questions and comments to Faye Butts fsb1@cornell.edu

This page was last modified on: 02/10/04