Printer Friendly

Experimental work on subsidies, moral hazard, and market power in agricultural markets.


Agricultural price support programs have existed as U. S. federal policy since the 1933 Agricultural Adjustment Act and most recently have been continued in the 2008 Farm Bill (USDA/ERS 2008). (1) Through the years one goal has been consistent: maintain the income of growers. Besides commodity-specific price supports, programs instituting supply controls, conservation programs, and marketing loans have been introduced in a succession of U.S. farm bills (Dimitri, Effland, and Conklin 2005). For decades, however, price supports have been the primary method of subsidization for commodity producers. These "coupled" subsidies, which are tied to price and production levels, are being scrutinized because of their potential trade-distorting impacts, resulting in an increased interest in new subsidies which are "decoupled" from production.

In this paper, laboratory markets are used to study both coupled and decoupled payment programs for agricultural commodities. We create a very simple and stylized subsidy environment that mimics how growers are subsidized in the United States. A target-price program guarantees every seller a price for each unit of a commodity brought to market. The guaranteed price is a promise that producers will receive no less than the target price for their output. Operationally, producers sell their output in the open market at an acceptable price. If the average market price is below the target price, the producer is given a deficiency payment equal to the difference between the target and the average market price for each unit sold. If the average market price is above the target or floor, the producer does not receive a subsidy. In addition to this coupled subsidy program, two decoupled payment schemes are also studied in this paper. One consists of a lump-sum payment made before there is any commodity production (2) and the other is a periodic payment made at the beginning of each new planting season. Each of the subsidy plans is not directly tied to price or production.

We construct two distinct market environments to study the impact of coupled and decoupled subsidies for agricultural commodities. Both are a two-stage game involving a production decision and then exchange. One market environment puts sellers in a competitive market, where they individually make production decisions. Collective production quantities then sell at a price (based on a demand schedule) that clears the market. The experimenter serves as auctioneer and buyer in this case. We refer to this as a "seller-only" institution. This experiment isolates the impact of subsidies on production decisions and ultimately market supply. In a perfectly competitive market, we would expect sellers to continue producing until marginal cost is just equal to the target price or the market price, whichever is higher.

The second market environment is more complicated. As in the previous market design, a production decision is made by individual sellers. But then their collective production is sold to buyers in a posted-bid auction. Sellers accept bids for offered prices and quantities until all units are sold. This environment is designed to capture potential impacts of buyer and seller interaction in price discovery along with production impacts of subsidies.

In all of our experimental sessions there are many trading periods or seasons. During each trading period there is a production decision followed by exchange. We conduct a baseline treatment without a subsidy, two treatments with generous and not-so-generous target-price supports, and two decoupled subsidy treatments with equivalent lump-sum and periodic payments.

If exchange is conducted through a posted-bid auction, a type of auction commonly used in agricultural markets, we show that generous or high price supports, as defined below, give buyers substantial market power. The program effectively results in helping commodity buyers get their bids accepted at low prices. Producers agree to lower prices for their output in this setting; hence, the market institution and the subsidy program create a moral hazard on the seller side of the market. This coupled subsidy program transfers healthy amounts of income to both the seller and the buyer. The subsidy cost to the government on each unit sold--the difference between the guaranteed and bid price--is high. We emphasize that this outcome is observed when the subsidy is generous. In our experiments, the amount of the generous subsidy is a target price near what producers would get if they were organized as a monopoly.

Not-so-generous price supports do not suffer from this problem. The target does not impact behavior if sellers have some probability of earning more than the guaranteed price. In this environment, producers work harder to get the best price for their output. We therefore tend to think of generous price supports as those at a level for which there is little or no probability of sellers exceeding in a market sale. Once sellers conclude that they cannot beat the target, moral hazard from the seller's side of the market enters the auction. Buyers then are able to get units at prices substantially below the intersection of supply and demand. Our experimental results show that in the face of generous price targets, market prices are below the predicted intersection of supply and demand by about 9 percent. Naturally, sellers produce more in both the seller-only and posted-bid auction institutions when there is a generous subsidy (13.9 percent and 6.8 percent more output, respectively), as compared to when there is no subsidy in each institution.

Our two other experimental treatments are motivated by policy discussions to decouple subsidy payments from crop price and production. One treatment pays a lump sum. When the experimental session begins, all producers/sellers receive a one-time, lump-sum payment. The second treatment is an "annual" payment made at the beginning of a "growing season" or trading period. Both payment plans are designed to give sellers income levels equivalent to accumulated target-price payments, as gauged by our target-price experiments with a generous target. These treatments are tested across both the seller-only and posted-bid auction environments. In our experiments, the lump-sum amount is paid above and beyond the regular participation fee that all subjects receive. In the annual-payment treatment, equal income payments are made to sellers at the beginning of each trading period and at the time a production decision is made. Again, the periodic payments are set to give sellers a similar income to that paid from the high price support. It is important to note that it is common knowledge to all agents in the subsidy experiments when added income is given to sellers. When subsidy payments are no longer directly linked to price or production, they are expected to have less impact on direct market outcomes. There are possible secondary impacts. The capitalization of decoupled payments into input values could, for example, impact related resource markets.


A. Market Basics

The subject pool was undergraduate university students recruited primarily from junior, senior, and graduate classes. (3) Each subject was seated at a linked personal computer station. Students made production and exchange decisions using a point-and-click screen format. Experimental sessions began with a standard presentation of instructions followed by one or more practice sessions. The practice sessions used costs and redemption values different from those in the actual experimental sessions.

The commodity being traded was referred to as a "unit." The unit redemption values and unit costs in Table 1 constitute individual induced demand and induced supply, respectively, for each trading period in the experiments. Summing the individual demand and supply relationships horizontally results in the induced market demand and induced market supply relations represented in Figure 1. This figure was not provided to subjects. For four sellers and four buyers, the predicted equilibrium price is 80 tokens and the equilibrium quantity is between 20 and 24 units. We estimate the inverse demand schedule as the linear form P = 135 - 10Q and the inverse supply schedule as P = 25 + 10Q, where P is price in tokens and Q is the quantity of units. For these functions, the Cournot solution for four sellers is a price of 86.10 and sales of 19.56 units. If sellers are perfectly collusive, the price is 98.34 and sales are 14.67. The Cournot solution for four buyers is a price of 73.90 and sales of 19.56. If buyers are perfectly collusive the price is 60.00 and sales are 16.00 units. Trading of partial units is not possible in our experiments; these predictions are used as benchmarks in the analysis below and in our discussions describing behavior.

Unit Buyer Redemption Values and Seller Costs (tokens) Used in the

Unit Redemption Value Cost

1 130 30
2 120 40
3 110 50
4 100 60
5 90 70
6 80 80
7 70 90
8 60 100

Before trading could begin there was production. Each seller decided how many units to bring to market, and the cost of production was assessed before units were sold. Thus, sellers held inventory entering the market, reflecting spot delivery of units. Inventory could not be earned over to the next period. In other words, the commodity could not be stored. The production decision corresponds to the beginning of a planting season. Once the planting decision is made, this is the amount of commodity that can be traded in the seasonal cycle. Producers can adjust future production decisions in later cycles, and they do as they learn more about the market. In actual runs we observed losses due to excessive production toward the end of early exchange periods, but they did not persist as trading cycles were repeated.

Earnings were denoted in a monetarily convertible currency referred to as tokens (1 token = 1 cent) and participants were paid in cash at the end of the experiment. Each session consisted of 20 production-exchange periods. The number of periods was unknown to participants until the session ended. Each participant was given an initial token balance of 700 tokens (or $7.00) at the beginning of a session. This initial balance was necessary because sellers incurred production costs prior to being given the opportunity to realize earnings from sales. Subjects were paid based on their performance in the experimental market. Subject payments were in the order of $37-$50 for sessions that lasted about an hour and a half.

All sellers had the same unit costs, and the same unit costs were used in each treatment (Table 1). Subjects were not aware that everyone had the same costs. Sellers could produce up to eight units, and sell units, possibly in a bundle, beginning with the lowest cost units. Seller earnings equaled the sale price minus the unit cost so that they earned profit by trading above unit costs. Table 1 shows that buyers could purchase up to eight units and redeem their purchases for amounts beginning at 130 tokens. All buyers had the same redemption schedules but did not know this. Buyer earnings equaled the redemption value minus the purchase price. Earnings accumulated during the sequence of trading periods and were displayed on each person's computer screen at the end of each trading period, after which a new production/exchange period began.

B. Exchange Institutions

A seller-only institution consisted of a simplified market in which all production from four sellers was sold (Figure 1); the market price from the demand schedule cleared all inventory. Trading moved briskly in these experiments because as soon as the production decision was made the computer sold all inventory at the market-clearing price determined from the market demand schedule. Producer/sellers immediately saw their earnings on the screen and went to the next production decision. No buyers were recruited in this treatment.

The second exchange institution was an auction designed to capture the main features of many agricultural markets. We refer to this institution as a posted-bid auction. As in the seller-only institution, each trading period begins with four sellers each making a production decision. Then sellers move into an auction with four buyers who place price bids for offered quantities. (4) Every bid required a price and quantity package. Bids were presented to sellers in a list that put the highest price and corresponding quantity first. Buyers were free to change their bid price and/or quantity as a clock ticked down. Sellers could accept bids and sell up to as many units the buyer was willing to take, provided the number of units produced was not exceeded. Once a buyer's bid/quantity offer was filled, it was removed from the list. A buyer could enter a new bid/quantity, but could not take more than a capacity limit of eight units per buyer per period. As buyers made bids for price and number of units, and units were sold, bids and sales were continuously updated on everyone's screen for the 3-min trading period, or until all production was sold. In every trading period agents had a complete history of activity.

Because they could post both a price and quantity in the auction, buyers could be more strategic in their offers. An offer to take more inventory at slightly lower prices could improve earnings over an offer at a higher price for fewer units. Buyers rarely purchased all of their units at the same price, and as observed in many repeated auction environments, where numerous identical or similar units are sold in sequence, prices tended to fall as units were sold (Ashenfelter 1989).

A posted-bid auction institution is designed to simulate a spot market for many agricultural commodities. A grower holds and then sells a fixed quantity of wheat, corn, soybeans, etc. Buyers, for example grain elevators, have limited capacity to accept units of a crop and post a price at which they will buy. The amount that can be accepted is known by sellers. Once the elevator is full, the buyer no longer posts a price. Posted prices change frequently and sellers are informed of a range of prices and know when they change.

C. Policy Treatments

Our policy treatments, designed to investigate the behavioral response associated with target-price programs in seller-only and posted-bid auction institutions, are (1) seller-only trading with no policy, (2) seller-only trading with a target price of 90 tokens, (3) posted-bid auction-trading and no policy, (4) posted-bid auction-trading and a target price of 84 tokens, (5) posted-bid auction-trading with a target price of 90 tokens, (6) posted-bid auction trading with a single decoupled lump-sum payment, and (7) posted-bid auction trading with periodic decoupled payments. The number of replications (rep) done for each treatment is reported in Table 2. More detail on how the policies were implemented in our experiments follows:
Estimated Base Convergence Levels and Treatment Adjustment Coefficients
(standard errors) for Alternative Trading Institutions and Policy

 Market Outcomes

Trading Institution/ Policy Treatment rep Price Trades

Seller Only No Policy 3 81.26 21.85
 (Base) (0.49) (0.22)
Posted Bid/ No Policy 5 1.33 *, (b) -1.43 *, (b)
 (0.56) (0.27)
Seller Only/Target 3 -5.89 *, (c) 3.03 *, (c)
 Price = 90 (0.56) (0.28)
Posted Bid/Target 9 -8.08 *, (d) -0.06 (d)
 Price = 90 (0.48) (0.20)
Posted Bid/Target 3 0.08 (be) -1.21 *, (be)
 Price = 84 (0.78) (0.29)
Posted Bid/Annual 3 -1.27 (ef) -2.90 *, (f)
 Payment (0.63) (0.22)
Posted Bid/Lump-sum 3 -1.40 (f) 1.32 *, (be)
 Payment (0.95) (0.21)

 Market Outcomes

 Seller Earnings

Trading Institution/ Policy Treatment Market Subsidized

Seller Only No Policy 154.51 154.17
 (Base) (2.16) (2.16)
Posted Bid/ No Policy 7.76 *, (b) 8.01 *, (b)
 (2.59) (2.60)
Seller Only/Target -36.02 *, (c) 52.94 *, (c)
 Price = 90 (2.53) (2.16)
Posted Bid/Target -41.97 *, (d) 49.62 *, (d)
 Price = 90 (2.01) (2.59)
Posted Bid/Target 0.51 (be) 15.61 *, (e)
 Price = 84 (4.00) (2.30)
Posted Bid/Annual -12.51 *, (f) 95.06 *, (f)
 Payment (3.31) (3.30)
Posted Bid/Lump-sum -6.00 (ef) N/A
 Payment (4.20)

 Market Outcomes

Trading Institution/ Policy Treatment Buyer Total
 Earnings (a) Market

Seller Only No Policy 142.81 1189.26 **
 (Base) (2.50) (2.94)
Posted Bid/ No Policy -8.89 *, (b) -2.65 (b), **
 (2.82) (4.98)
Seller Only/Target 34.95 *, (c) -7.73 (bcd), **
 Price = 90 (3.10) (4.53)
Posted Bid/Target 31.17 *, (c) -38.15 (e),*, **
 Price = 90 (2.26) (4.99)
Posted Bid/Target -5.13 (bd) -9.00 (bcf)
 Price = 84 (2.91) (5.87)
Posted Bid/Annual -6.55 (bd) -74.60 *, (g)
 Payment (3.61) (5.40)
Posted Bid/Lump-sum 3.83 (e) -4.52 (bdf)
 Payment (4.46) (3.36)

* Indicates that the convergence level for this policy treatment is
significantly different from the base treatment, 95% confidence level.
** Indicates that the convergence levels for seller and buyer market
earnings are significantly different, 95% confidence level.
(a) Buyer earnings for seller-only markets are calculated from average
revenue and average cost.
(b-g) Same letter indicates no significant difference between the
convergence levels in the respective equations. Different letters
indicate a significant difference between convergence levels, 95%
confidence level.

No Policy. A base treatment for both exchange institutions (seller only and posted bid) consisted of a market in which no policy is implemented and allows for comparison of how the market might be impacted under a subsidy policy.

Coupled Payments. A target-price policy treatment provides deficiency payments paid to producers equal to the positive difference between the target price (TP) and the average market price (MP) times the number of units sold by the individual seller. That is, (TP - MP) x (Units Sold). Sellers and buyers are made aware of this feature via the experiment instructions prior to conducting the experiment. Specifically, they were informed in their instructions that:

* There is an additional payout to sellers that will be added to this experiment--a deficiency payment.

* If the average market price for all units sold during a period is below a target price of--tokens, there will be a deficiency payment made to all sellers on all units sold, regardless of price the seller received.

Two target-price programs were put in place: one that paid a high price guarantee of 90 tokens and another that paid a relatively low guarantee of 84 tokens. The high target price (90 tokens) was calculated from the predicted equilibrium price (80 tokens) plus a premium calculated using actual historical U.S. commodity target prices for wheat. The target price of 84 was based on the asymptote estimated from a convergence model (explained later) using data from initial posted-bid runs. (5)

For each target-price treatment more details were provided to participants, for example, for the 90-token guarantee the instruction read further as follows:

Deficiency Payment.

* The target price = 90 tokens.

* DP = (90 - average market price) x units sold by each seller.

* If the average market price is greater than the target price, no deficiency payment is made.

* The amount of the DP, if one is made, will be on the period recap screen and will be added to the total earnings.

A target price of 90 is above the seller Cournot amount of 86.10 tokens and below the perfectly collusive price of 98.34; hence, the target price reflects a level of cooperation that is better than Cournot and would probably require communication among sellers to achieve absent the subsidy program. A target price of 84, however, is just below the Cournot solution. This price is achievable for four sellers, if they are able to recognize that from the competitive solution each seller can unilaterally reduce production and increase profits.

Decoupled Payments. Experimental markets are well suited to study decoupled subsidies. Experimenters almost always give subjects a participation fee, independent of later decisions they make. Our experimental instructions show that every subject received 700 tokens at the beginning of the experiment. It is a relatively easy matter to extend this fee to a per-session or a period-by-period payment at the beginning of each trading period. The payment will be asymmetric. It will be common knowledge that sellers receive more than buyers. In policy circles, boosting the participation fee of sellers is giving the farmer a lump-sum bond. Making periodic payments to sellers independent of production and trading decisions is providing a proposed annual bond.

To implement decoupling with the posted-bid auction institution and a lump-sum payment, we increase the participation fee paid to each seller, but not the buyers. Each seller is now given 2,140 tokens at the beginning of a session. The annual-payment treatment does not change the participation fee, but at the beginning of each trading period, sellers are paid 107 tokens for each of the 20 trading periods in a session. (6)


The data were analyzed in two ways, graphically and statistically through results from a standard convergence model. The latter is detailed below. Our description and interpretation of results from the experiments will rely on both the convergence estimates and observed relations from the graphs.

A. Convergence Model

Data collected in the seller-only experimental market included price, quantities traded/produced, and earnings. Variations of the following general convergence model (1) were estimated to describe the data and allow for statistical comparisons (Ashenfelter and Genesove 1992; Noussair, Plott. and Riezman 1995):

[] = [B.sub.0][(t - 1)/t] + [B.sub.1](1/t) + [[i-1].summation over (j=1)][[alpha].sub.j][D.sub.j][(t - 1)/t] + [[i-1].summation over (j=1)][[GAMMA].sub.j][D.sub.j](1/t) + [], (1)

where [] = average sale price (or units traded or produced and earnings) across the three replications and all trades for each of t periods in cross section (or treatment)i; [B.sub.0] = the predicted asymptote of the dependent variable for the base treatment, which is the seller-only, no-policy treatment; [B.sub.1] = predicted starting level of the data for the base treatment; t = trading period 1, ..., 20; [D.sub.j] = dummy variable representing the jth set of test treatments; and [] = error term. As the focus of this series of experiments is the impact of differing policy treatments, we are primarily interested in differences between the asymptotes for each test treatment.

The Parks (1967) method was used to estimate the model. This is an autoregressive model in which the random errors, [], have structures E([]) = [[sigma].sub.ii] (heteroscedasticity); E([] [u.sub.jt]) = [[sigma].sub.ij] (contemporaneously correlated); and [], = [[rho].sub.i][u.sub.[i,t-1]] + [[member of]] (autocorrelation). The Parks method assumes a first-order autoregressive error structure with contemporaneous correlation between cross sections. The covariance matrix is obtained by a two-stage procedure leading to the estimation of model parameters by generalized least squares. See SAS (2008) for details of this estimation method. The use of the Parks method takes into account the unique statistical problems resulting from the panel data sets that consist of time series observations on each of the several cross-sectional units generated in our experiments and allows for simultaneous testing across treatments. Analysis was conducted in SAS using the TSCSREG Procedure. (7) Differences between convergence levels are considered to be significant at the 95 percent confidence level.


A. Subsidies Coupled to Price and Production

Graphical and statistical analyses of market outcomes (prices, trades or production, seller market and subsidized earnings, buyer earnings, and total earnings) are reported below for posted-bid auction and seller-only trading institutions with no policy and the two target-price policies. For convergence model analysis, the seller-only institution with no policy in effect was used as a base treatment. Results are averages across replications.

We begin by focusing on the seller-only prices in Figure 2. With no target price, sellers collectively produce enough units to result in a market price near the competitive equilibrium of 80 tokens. Average price is exactly at 80 for about half of the trading periods. There is a tendency in other periods for the prices to be slightly higher, and the convergence level reported in Table 2 is 81.26. With a guaranteed target price of 90 in the seller-only experimental design, average market prices drop and stabilize at about 77 tokens in the figure. The estimated asymptote calculated from coefficients in Table 2 is 75.38. The market price falls because sellers produce more to collect the subsidy associated with the 90-token target price on each unit sold. Table 1 shows that each subject should produce 6 units for a total of 24 in the market, and as reported in Figure 3 this is what they do. The convergence quantity is in this neighborhood.



In the freely operating posted-bid auction, labeled in the figures and tables as "Posted Bid/No Policy," there is a slight rise in the observed price level; all but one of the trading periods show average prices above 80 tokens, and the estimated convergence price is 82.60 tokens. We attribute this higher price to risk aversion from excess inventories that could sell below cost, a result consistent with previous work (Phillips, Menkhaus, and Krogmeier 2001). Sellers produce about 20 units in each trading period, which is the minimum of the competitive range of 20-24 units. If we examine average trade prices period by period in Figure 2, we observe prices hitting near 83 tokens on a regular basis. The seller Cournot solution is 86.10 tokens. Sellers in this institutional setting are able to approach the Cournot price, and it may be that aversion to inventory losses is a facilitating influence toward reaching this outcome. As sellers move toward the Cournot price solution, they all realize higher profits, and have no incentive to individually produce more. We therefore consider behavior to be stable.

Our not-so-generous coupled subsidy sets a target price of 84 tokens in the posted-bid auction; we see from the convergence estimates in Table 2 that this price guarantee has virtually no impact on behavior, relative to no policy in a posted-bid auction. In particular, quantities produced continue near 20 units in each trading period. Prices are lower in the first part of the experiment, but after period 14 Figure 2 shows that they are tracking closely with the no-policy prices. The target price of 84 is below the Cournot price, which we have already noted is achievable by the four sellers. We think it is important to observe in Figure 2 that average trade prices begin at about 80 tokens, and sellers are given a deficiency payment of about 4 tokens per unit during these trading periods. It then seems to us that sellers realize that they can do at least as well in the market. Prices steadily increase toward the Cournot level and deficiency payments are unnecessary at the end of the sessions. The target price as a safety net is in place at a level above the equilibrium, but close enough to the Cournot price that sellers can beat the target.

A generous target price of 90 tokens is between the Cournot and monopoly price of 98.34. We observe a sharp change in seller behavior from this policy. It is most noticeable in Figure 2. The average price per unit drops to about 73 tokens in a trading period. From Table 2, 73.18 is the estimated convergence level. A price of 73 is slightly below the buyer Cournot price of 73.90 and above the buyer collusive (monopsony) price of 60.

This pricing behavior leads us to the conclusion that sellers are willing to accept any price in the auction market. With a target price of 90 there is no incentive to allow buyers to compete for units. Consequently, Figure 2 shows that prices immediately fall in favor of buyers. The Cournot quantity for buyers is about 20. Actual sales are estimated at about 22 units per trading period. Our sense of the market is that since sellers are guaranteed a price of 90 for every unit sold, they give way to buyers in the auction, who operate close to their Cournot market price.

Graphical representations of average seller subsidized earnings are presented in Figure 4. As expected, subsidized seller earnings with a target price of 90 are highest. For both the seller-only and the posted-bid treatments, sellers are earning about 200 tokens per period from the subsidy. The posted-bid auction institution is as good as a stylized seller-only market institution at delivering income to the seller. When the target price is 84, market earnings are close to the earnings from the seller-only no-policy treatment and not statistically different from those in the posted-bid market with no policy. The posted-bid auction subsidized seller earnings with a target price of 84 are near the Cournot prediction of 171.03. (8) Table 2 estimates subsidized seller earnings for the target price = 84 treatment at 169.78 tokens. Thus, a subsidy payment that goes from 84 to 90 tokens per unit of commodity gives the seller about 30 more tokens of income (about 17.6 percent) per trading period.


The target price = 84 safety net program costs the experimenter almost nothing in guarantees, because it is set near a level achievable in the posted-bid environment. The target price = 90 program is costing the experimenter about (90 - 73) x 22 = 374 tokens each trading period. This is a subsidy payment of 93.5 tokens for each seller. In an auction environment, the experimenter is paying about 93.5 tokens each trading period to get a seller from an earnings level of about 162 tokens to about 200 tokens.

Buyers benefit substantially from high subsidies. Sellers still benefit the most, receiving 17.6 percent or 30 tokens per period compared to 8.7 percent or 12 tokens going to the buyer. Before the subsidy in the posted-bid auction, buyers purchased about 20.41 units in a trading period at a price near 82.60 tokens per unit; total payments were 1,652 and total redemption values from Table 1 were 550 per buyer or 2,200 for the four buyers. The four buyers earned about 550 tokens each period or 138 tokens per buyer. After the subsidy, prices fell to 73 per unit; about 22 units were sold, so total payments fell to 1,600 tokens and earnings increased to 600 tokens or 150 tokens per buyer. The subsidy increases buyer earnings by about 12 tokens or 8.7 percent.

Results from the coupled target-price subsidy treatments are not surprising once the bargaining behavior of sellers is understood. These subsidies create moral hazard in an auction in which buyers post bids. When the high subsidy is in place, sellers simply want to produce as much as they can, get the units sold, and collect the target-price payment. They have no incentive to make the auction institution work for them as they do without a subsidy or when the coupled subsidy is near their Cournot solution. Consequently, buyers have their way in the auction and prices are close to their Cournot solution.

B. Decoupling the Subsidy

The above results indicate that the target-price policy is a relatively inefficient income transfer mechanism. Policymakers are interested in ways to deliver income support to sellers without impacting market outcomes--prices and production or trades--i.e., a "decoupled subsidy" (Sumner 2005). What about giving sellers more tokens at the outset of the experimental session? This is like paying producers a Jump sum when they make the long-run decision to enter, or remain in, the industry. A related way to deliver extra income to the producer is to simply pay the seller a portion of the sum at the beginning of each trading period. This is like giving the agricultural producer an annual income payment independent of what that producer decides to do in the market. Both of these payment methods fall under policy discussion of decoupling the subsidy from commodity price and production.

Given the measured impacts of using target prices and the resulting cost of delivering income to sellers in our experimental markets, there is merit to simply giving sellers payments independent of the market. We are aware, however, of equity issues that arise between bargainers when there are different levels of wealth. Buyer's knowledge of sellers earning additional income could impact their behavior, i.e., buyers may try to capture some of this subsidized surplus during price negotiations. Parties tend to bargain toward equality of wealth. (9) Hence, giving sellers a relatively better wealth position than buyers may create moral hazard in trading, just as the high target price did.

Figures 5 and 6 show the results of these treatments. Average prices and trades (or production) from the no policy and target price = 90 policy treatments also are included for comparison. Estimated convergence market outcomes for the treatments described above are included in Table 2.



The lump-sum and periodic payments appear to have only a marginal impact on prices, compared to the no-policy treatment. Prices under these two treatments, in fact, track the competitive prediction in the market better than the market with no policy in place, which we earlier described as the result of Cournot behavior. Trades range from 19 units in the annual-payment treatment to about 22 units in the target price = 90 treatment. There is no difference in trades/production between the no-policy and lump-sum treatments. Results suggest that decoupled subsidies do not impact participant expectations about market prices as is the case with the target-price policies. Thus, when making production decisions, i.e., when equating price to marginal cost, production is not distorted under decoupled policy treatments. The specific target price given also may impact expectations about price. As floor prices are not set in decoupled treatments, marginal relationships are not impacted by these expectations under decoupled policy. For policymakers looking for ways to increase the income of farmers without affecting incentives, these results are encouraging. From a market efficiency standpoint, the lump-sum payment seems superior, but it may be more costly to implement than an annual-payment scheme.

C. Efficiency Considerations

Table 2 reports convergence model results related to earnings--buyer earnings, sellers' subsidized and unsubsidized market earnings, and total market earnings. At the predicted equilibrium, total potential market surplus or earnings (from Figure 1) is 1,200 tokens. If potential surplus were divided equally amongst buyers and sellers at this equilibrium, individuals would receive 150 tokens per period each. Agents in the seller-only and posted-bid environments are relatively efficient at extracting the total potential surplus with no policy in place (total earnings are 1,189.26 and 1,186.61 tokens, respectively). The posted-bid lump-sum treatment is nearly as efficient with an estimated convergence of 1,184.74 tokens. As discussed previously, estimated buyer earnings are highest with a target price of 90. Total earnings of 1,114.66 tokens suggest that deadweight loss or inefficiency is greatest in the posted-bid environment in the annual-payment treatment. Compared to the lump sum, sellers make about 4.22 less trades per period with the annual payment. The inefficiency associated with the generous target price is largely related to inefficient pricing.

Figure 6 shows that sellers in the annual-payment treatment consistently produce at lower levels than sellers in the other treatments reducing overall market efficiency. The convergence model estimates sales at 18.94 units while the other treatments are showing trades of about 20-24 units. Interestingly, the lump-sum payment scheme is not showing this effect. Here estimated production is 20.53 units per trading period, no different than the posted-bid no-policy treatment. This decline in production is a cause for concern if annual bond payments should become practice.


The results of this research suggest that the target-price policy creates a moral hazard in the posted-bid auction environment. With a "high" subsidy in place, sellers have no incentive to make the trading institution work for them and produce more. As a result, buyers move the auction price close to their Cournot solution.

Buyers benefit more from a target-price policy than from decoupled lump-sum or periodic subsidies. Decoupled income enhancement policies better achieve the goals of improving producer incomes, while not distorting market signals. In international commodity markets, distortion is becoming an important trade issue. World Trade Organization negotiations, for instance, "have placed a premium on a practical understanding of the magnitude of production effects of the whole range of farm programs" (Sumner 2005, p. 1229). Our experimental results show when target prices are most likely to cause distortion and measure the distortion for a set of demand and supply conditions. Overall, these results suggest that there would likely be resistance to decoupled payments by agribusiness firms or "buyers" of agricultural commodities.

A related issue, but one not specifically addressed in this study, is how alternative policies impact resource (land and labor) markets. It is common knowledge that agricultural subsidies, such as a target-price deficiency payment, are capitalized in land prices. But, what are the effects of a decoupled subsidy on land values and rental rates? It is possible that the less transparent the connection of the payment to crop price/production is, the less capitalized the subsidy will be. It may be that if lump-sum payments are tied to the producer's household income, then the size of the payment is determined by factors other than just the productivity of the land. Further study on making bond payments operational is in order. The aim should be to minimize their impact on land values, along with decoupling payments from prices and production decisions.

Further, the income subsidies (lump sum and annual) in our last two experimental policy treatments may damage labor incentives, in the sense that as income increases less labor is supplied to production. Fewer units sold at or near the competitive price would reflect a lack of effort in our experiments. There is an indication of this occurring in the annual-payment treatment.

As policymakers investigate income transfer mechanisms for agricultural producers, there is a paucity of analyses comparing new alternatives to policies currently in place. Experimental methods seem uniquely suited to such ex ante analyses. These results suggest that current deficiency payment policies that have a higher subsidy level may create a moral hazard problem when producers market their products, and therefore these policies are relatively inefficient transfer mechanisms. Decoupled mechanisms may better achieve the objective of income transfer while creating fewer distortions. However, political resistance to these policy mechanisms seems likely from buyers and potentially sellers. More research investigating potential impacts on other resource markets is certainly warranted.


Ashenfelter, O. "How Auctions Work for Wine and Art." Journal of Economic Perspectives, 3(3), 1989, 23-36.

Ashenfelter, O., and D. Genesove. "Testing for Price Anomalies in Real-Estate Auctions." American Economic Review, 80(2), 1992, 501-505.

Davis, D. D., and C. A. Holt. Experimental Economics. Princeton, NJ: Princeton University Press, 1993.

Dimitri, C., Effland, A., and N. Conklin. "The 20th Century Transformation of U.S. Agriculture and Farm Policy." Economic Information Bulletin No. 3, USDA/ERS, Washington, DC, 2005.

Friedman, D., and S. Sunder. Experimental Methods--A Primer for Economists. Princeton, NJ: Princeton University Press, 1994.

Goodwin, B. K., and A. K. Mishra. "Another Look at Decoupling: Additional Evidence on the Production Effects of Direct Payments." American Journal of Agricultural Economics, 87(5), 2005, 1200-1210.

Key, N., R. N. Lubowski, and M. J. Roberts. "Farm-Level Production Effects from Participation in Government Commodity Programs: Did the 1996 Federal Agricultural Improvement and Reform Act Make a Difference." American Journal of Agricultural Economics, 87(5), 2005, 1211-1219.

Noussair, C. N., C. R. Plott, and R. G. Riezman. "An Experimental Investigation of the Patterns of International Trade." American Economic Review, 85, 1995, 426-491.

Parks, R. W. "Efficient Estimation of a System of Regression Equations When Disturbances are Both Serially and Contemporaneously Correlated." Journal of the American Statistical Association, 62, 1967, 500-509.

Phillips, O. R., D. J. Menkhaus, and J. L. Krogmeier "Laboratory Behavior in Spot and Forward Markets." Experimental Economics, 4, 2001, 243-256.

Plott, C. R. "Industrial Organization Theory and Experimental Markets." Journal of Economic Literature, 20, 1982, 1485-1527.

Roth, A. E. "Bargaining Experiments,", Chapter 4 in The Handbook of Experimental Economics, edited by J.H. Kagel, and A. E. Roth. Princeton, NJ: Princeton University Press, 1995.

Roth, A. E., M. W. K. Malouf, and J. K. Murnighan. "Sociological versus Strategic Factors in Bargaining." Journal of Economic Behavior and Organization, 2(2), 1981, 153-177.

SAS. "The TSCSREG Procedure / Parks Method (Autoregressive Model)," SASOnlineDoc 9.1.3.,

Smith, V. L., and A. W. Williams. "The Boundaries of Competitive Price Convergence," in Bargaining and Market Behavior: Essays in Experimental Economics, edited by V. L. Smith. Cambridge, England: Cambridge University Press, 2000.

Suits, D. B. "Agriculture," in The Structure of American Industry, 11th ed., edited by W. Adams and J. Brock. Upper Saddle River, NJ: Prentice Hall, 2005, 1-22.

Sumner, D. A. "Production and Trade Effects of Farm Subsidies: A Discussion." American Journal of Agricultural Economics, 87(5), 2005, 1229-1230.

United States Department of Agriculture Economic Research Service. (2008), "2008 Farm Bill Overview,"

United States Department of Agriculture Farm Service Agency and National Statistical Service. Wheal: Farm Prices, Support Prices and Ending Stocks, 1955 to 2004. 2004.

(1.) For detailed discussions on the history and extent of support programs and empirical testing of their impacts with data, e.g., from the U.S. Agricultural Census and Agricultural Resource Management Survey, see Suits (2005), Goodwin and Mishra (2005), Key, Lubowski, and Roberts (2005), and Sumner (2005).

(2.) In the experiment design, the lump-sum payment is assumed to be made to a grower at the inception of this subsidy program.

(3.) Experimental studies have traditionally used students as subjects due to convenience, low opportunity costs, relatively steep learning curves, and some lack of exposure to confounding external information (Friedman and Sunder 1994, p. 39). The advantages of using student subjects often offset potential disadvantages.

(4.) We find that four sellers and four buyers in an auction environment is ideal for studying market behavior (Phillips. Menkhaus, and Krogmeier 2001). Previous experimental market work has shown four sellers to be adequate. Plott (1982) finds that in offer markets "experiments with three and four sellers converge with regularity to the competitive equilibrium" (p. 1496). Smith and Williams (2000) cite "a substantial body of evidence suggests that markets organized under double-auction trading rules converge 'rapidly' to a CE [competitive equilibrium] price when there are as few as four sellers and four buyers" (p. 289). They find that auction convergence is robust in a duopoly: "we conclude that the CE model appears to provide a satisfactory prediction of actual market outcomes with as few as two sellers" (p. 302). Davis and Holt (1993, p. 181) conclude that posted auctions move toward the predicted equilibrium and achieve efficiency comparable to the double auction. Together these findings provide the impetus for us to use four buyers and four sellers in our posted-bid auction.

(5.) Target-price subsidy calculations go as follows:

Target price = 90 tokens.

The predicted competitive equilibrium price of 80 tokens (see Figure 1) was used as a starting point. A premium of 13.8 percent was calculated using U.S. wheat prices and target prices between 1974/75 and 1995/96 (Source: Farm Service Agency and National Agricultural Statistical Service, USDA, Wheat: Farm Prices, Support Prices and Ending Stocks, 1955-2004):

(Average 15 yr Target Price)/(Average 15 yr Wheat Price Received) = Premium. Hence ($3.71/bu)/($3.26/bu) = 1.138 or 13.8 percent.

The competitive equilibrium price of 80 increased by 13.8 percent yields an approximate target price of 90.

Target price = 84 tokens.

Prices from the posted-bid auction with no policy converged to 83.74 or about 84 tokens. Only price was considered in the convergence model.

(6.) Decoupled subsidy amounts were calculated to equal the average payment per seller paid out for 20 trading periods under the target-price treatment with a target price of 90 tokens outlined above for the auction-trading institution--from the initial three replications.

(7.) The convergence model describes the path of the data and provides a means to compare treatment effects by testing statistical differences across predicted treatment asymptotes for each market outcome. Other approaches, such as random or fixed effects models, could have been used, but these are often more appropriate for causal models rather than data description.

(8.) The seller Cournot price is 86.10 and the market quantity is 19.56 tokens. Total Cournot revenue is 1,684.11 tokens. Each seller will produce to 5 units on their cost table. For four sellers total cost is 1,000 tokens. Total earnings are therefore 684.11 or 171.03 tokens per seller.

(9.) Roth in Kagel and Roth (1995) makes the point that in all kinds of contexts people bargain toward 50/50 splits. Roth et al. (1981) find that bargaining behavior is strikingly different in binary bargaining games, depending on the information given to all players about payoffs, or that given to just one player about another player's payoffs. Players however tend to bargain toward equal gains.


DP: Deficiency Payment

MP: Market Price

TP: Target Price

* Support from the Paul Lowham Research Fund is gratefully acknowledged. This research was also supported under the United States Department of Agriculture. Economic Research Service/University of Wyoming Cooperative Agreement--"A New Generation of Farm Policy Tools: Identifying and Assessing Economic Implications" (USDA ERS N45104). Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the funding sources.

Phillips: Department of Economics & Finance, University of Wyoming, Laramie, WY 82071-3985. Phone 1-307-766-2195, Fax 1-307-766-5090, E-mail

Nagler: Department of Agricultural & Applied Economics, University of Wyoming, Laramie, WY 82071-3354. Phone 1-307-766-5615, Fax 307-766-5544, E-mail

Menkhaus: Department of Agricultural & Applied Economics, University of Wyoming, Laramie, WY 82071-3354. Phone 1-307-766-5128, Fax 1-307-766-5544, E-mail

Bastion: Department of Agricultural & Applied Economics, University of Wyoming, Laramie, WY 82071-3354. Phone 1-307-766-4377, Fax 1-307-766-5544, E-mail


COPYRIGHT 2010 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Phillips, Owen R.; Nagler, Amy M.; Menkhaus, Dale J.; Bastian, Christopher T.
Publication:Contemporary Economic Policy
Article Type:Report
Geographic Code:1USA
Date:Oct 1, 2010
Previous Article:The impact of antismoking policies in Korea on quit success and smoking intentions.
Next Article:How does globalization affect Industrial competitiveness?

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters