Market integration and transaction costs

The fact that the existence of transaction costs keeps many rural households from participating in certain agricultural markets has been documented in the economics literature by De Janvry, et al. (1991). Transaction costs drive wedges between purchasing and selling prices of a household, based on the concept of non-tradable goods taken from international trade theory.

However, the literature has not used the same concept to determine why one household opts for a particular sales market for its product while another does not. Although risk considerations obviously could determine that a household will diversify the markets for its product, the transaction costs associated with each household and the differential transaction costs between markets would also help explain the "mix" of destinations a farmer chooses.

We have slightly modified the methodology proposed by De Janvry, et al. (1995) in two aspects to account for the direct measurement of transaction cost. First we are modeling the decision of selling at the farmgate or selling at market. We believe that the decision of a

3 This can be done using the «hedonic price» technique. See Section 5.2.3.

household to participate in a certain agricultural market depends on that household’s position of supply and demand relative to the range of prices created as a result of the difference existing between effective buying and selling prices on that market. This range originates from a group of transaction costs, some of which are specific to the household, while others are related to the environment or region in which the household is located and still others are associated with the specific market of destination.

In this context, a particular market "fails" when a household is faced with a large difference between the price at which a product or input could be bought and the price at which it could be sold. Given the wide margin between these two prices, it may be better for the household not trading the product or input on that market. While this decision occurs in all markets to which the household is associated, the household will prefer to remain self-subsistent for that crop.⁴ Generally, households can be classified in different categories according to the

"mix" of markets in which they have decided to participate.

The second modification, which will be described in more detail in the next section, is the introduction of a hedonic price function to account for the transaction costs differences.

If p is the effective price that determines production and consumption decisions, each household faces the following:

Supply of agricultural product (4)

Demand of agricultural product in market j (5)

Idiosyncratic transmission of prices in market j (6)

Transaction costs in market j (7)

where z^q, z^dj, z^pj and z^ij are exogenous variables that affect supply, demand, sales price and transaction costs, respectively. Thus, for the retailers of a product in market j, the effective price at level of each household would be:

(8) In this framework, the condition of being a retailer of potato in market j would be:

(9) This model can be estimated using the following probit equation:

(10)

4 In this case, the shadow or subjective price of the household (that which equals its supply and demand) falls within the margin: it is higher than the price the farmer would receive if he had sold the product, for which reason he decides not to sell; and is lower than what it would cost him to buy the product, for which reason he decides not to buy it.

The expanded model can make estimates based on either a probit or logit specification or a multivariate probit or logit, depending on whether we are dealing with two or more destinations. If we use the participation of sales in each market as the base and take into account that the endogenous variable is between values 0 and 1, the valid estimation method would be a Two-Limit Tobit Model. In our case, we are attempting to simulate a strategy associated with the decision to sell at the farmgate or elsewhere so we will try to capture this decision using a probit model.

5.3.2 Strategies used to measure transaction costs

After estimating the equation (10), the reduced form of the equation of supply conditioned on the selected strategy can be derived:

(11) The estimation of equation 11 equals an estimation in two stages, where the Mills ratio is introduced [obtained from estimating equation (10)] to take into account the endogenous nature of the decision (sell only at the farmgate or also sell at other locations).

To associate transaction costs to the effective price each farmer receives, we chose to estimate a hedonic price equation. The word "hedonic" is normally used in the economics literature to refer to the underlying profit that is obtained when consuming a good or service.

A good that has several characteristics generates a number of hedonic services. Each one of these services could generate its own demand and would be associated with a hedonic price.

Rosen (1974) developed the theoretical framework on which hedonic models are based. We interpret the model somewhat differently. The price the farmer receives has a set of "premia"

or "discounts" for a series of services that have been generated, or perhaps omitted.

Therefore, the average farmgate price can be defined as a function of hedonic prices, which is simply the mathematical relationship between the prices received by this added value (i.e. potato) and the characteristics of the transaction associated with this product. This is:

(12) where P_j is the average price obtained by j-th farmer for the potato sale; and where (z_1j, z_2j ...z_Kj) represents the vector of characteristics associated with the transactions completed by the farmer. The price function was estimated in accordance with the strategy followed.

It is clear in the literature of hedonic price functions that h(z) does not strictly represent a "reduced form" of the functions of supply and demand that could be derived from the

5 See Rosen, S. (1974) or Wallace, N.E (1996)

production or utility functions of the economic agents involved in the transaction.⁵ Rather, h(.) should be seen as a restriction in the process of optimization of sellers and buyers. Rosen (1974), and more recently, Wallace (1996)showed that while growing marginal costs exist for some of the characteristics (in this case associated with the generation of information, negotiation and monitoring of the transactions) for farmers and/or sellers, the hedonic function could be non-linear. In this case, the non-linearity would mean that the relative importance of transaction costs is not the same for all farmers.

The estimation of an equation such as the one proposed here permits us to disaggregate the price received by the farmer into a series of components associated with the attributes of the transactions. A complementary way of interpreting this equation is where the constant estimate represents a price indicator that results from following the "law of one price", the rest of the equation being the elements that must be discounted from the price due to the differences in the distance of the farmers from the market and other associated transaction costs. Comparing the transaction costs between households with different endowments (private and public assets) will allow us to understand the importance of key assets in reducing transaction costs.

5.4 Transaction costs in rural Peru