by Bill Blake (UC Cooperative Extension)
November, 1994
Readers are invited to submit comments to Bill Blake <ceplacer@ucdavis.edu> or to OPENAIR-MARKET NET (mar@interaccess.com). Comments submitted to Openair-Market Net will be considered for publication on the web site.
A farmers' market is both an event and a point of sale. As an event, it provides a focal point for a community, an opportunity for direct exchange between growers and consumers, and a place for socializing. As a point of sale, according to conventional wisdom, it offers better quality produce at lower prices because farmers can market directly to consumers. Produce from farmers' markets has been shown to be preferable to produce from grocery stores (Sommer, et al., 1982), and this improved quality should tend to increase prices. The additional utility of participating in the "event" should also push prices higher. On the other hand, the market is less convenient than a grocery store, and the produce is not as clean, uniform, or cosmetically controlled. These factors exert a downward pressure on prices.
This paper tests the hypothesis that prices at farmers' markets are lower. Since there are forces working both to raise and lower prices, this is an empirical question. The question is interesting because once it has been determined whether prices are lower or higher, one can address the importance of the utility that consumers derive from the "event" of the market.
In previous work, Sommer, et al. (1980) looked at markets in Northern California. In a survey of nearly 358 items over the course of several months, they determined that prices were lower by 37% for vegetables and 39% for fruits. Their article also rehashes the result of divers other studies, mostly from the East Coast, which indicate savings ranging from 8% to 50%. They point out, however, that these other studies are not comparable, and that they suffer from ambiguity around the term "farmers' market." The present paper has the convenience of using the same definition as the Sommer, et al. survey, that of a market certified by the California State Department of Food and Agriculture.
While the Sommer, et al. survey was extensive and complete, it did not attempt a multivariate analysis. A great number of factors influence grocery store prices, and most likely also affect prices at markets. These include size of the store, amount of local competition, population, local median income, etc. (Cf. for example Maley, 1976). This study employs a multivariate regression analysis so that the influence of each of several variables could be tested and quantified.
However, the effect of marketing directly could not be separated from other aspects of the farmers' market. The market is simply a different product than the grocery store. Freshness, taste, and the experience of the event are all interior to the produce at a market. Convenience, sterility, cosmetic quality, and one-stop shopping are likewise interior to the grocery store's produce. This paper does not simply compare produce at one final purchase point with produce at another. It compares farmers'-market-produce to grocery-store-produce, with the many sociological and psychological attributes that both composite goods entail.
The analysis is conducted on a set of cross-sectional data collected by the author over an eight-day period in mid- February 1994 in five Northern California cities which have regular, year-long, certified farmers' markets. Those places were Davis, Sacramento, San Francisco, Berkeley and Marin. Prices were collected for seven widely available vegetables or fruits: greenleaf lettuce, red potatoes, yellow onions, bunched kale, Fugi apples, oranges, and kiwi fruits. Prices in each locale came from the farmers' market and from three neighboring stores, and were differentiated into certified organically grown and otherwise produced. An item could have a theoretical total of 26 observations (8 from markets and 18 from stores, or 19 conventional and 7 organic). Actual totals ranged from 18 to 25 observations.
Individual prices for each item were then divided by the sample mean for that item's prices in a single locale to create an indexed price:
zij= Pij/Pi
where zij = indexed price of ith produce at jth point of sale
Pij = ith price at jth point of sale
Pi = sample mean of an item's prices in a particular city
Data were collected for a number of independent variables. The variety (var) of produce was measured by counting the number of different produce items offered. As is often the case in dealing with produce, this led to problems of definition. In this survey as in Sommer, et al., 1980, if a seller made a distinction it was honored in the variety count. Thus, bunched beets and loose beets are two items, large oranges and extra large oranges are separate, and some stores had nearly a dozen apple varieties. Number (num) was meant to capture the effects of competition by measuring the number of competitors. For stores, managers were simply asked how many stores they competed with. For markets, this was defined as the number of stalls offering five or more different items. The number of shoppers (shop) per day controls for customer traffic. A higher volume of traffic can account for price differentials, both between stores and between the markets (some of whom have very high volumes, especially in the peak summer season) and grocery stores. The size of the selling area is measured in square footage, and is included to test the relationship between the physical size of the store or market and prices.
There is a dummy variable for organically grown (org) produce, 1 if yes, 0 otherwise. A second dummy variable records the effect of farmers' markets (mark) on prices. Six dummy variables, v1 to v6 capture any variability due specifically to any of the seven vegetables. A final variable questions whether a store being a co-operative (coop) has any effect on prices. Guy and O'Brien (1983) found a significant difference in prices between multiple retailers and co- operatives on the one hand, and affiliated and independent retailers on the other. In fact, a difference in prices for farmers' markets could be seen as simply a special case of the effect of corporate structure on prices.
The model to be estimated is given by:
(1) zij = zij(var, num, shop, size, org, mark, coop, v1-v6)
where:
var = number of different produce items available at store or market
num = number of competitors
shop = number of shoppers per day
size = square footage of market or store
org = dummy variable: 1 if organically grown, 0 otherwise
mark = dummy variable: 1 if a farmers' market, 0 otherwise
v1 = dummy variable: 1 if lettuce, 0 otherwise
v2 = dummy variable: 1 if potatoes, 0 otherwise
v3 = dummy variable: 1 if onions, 0 otherwise
v4 = dummy variable: 1 if kale, 0 otherwise
v5 = dummy variable: 1 if oranges, 0 otherwise
v6 = dummy variable: 1 if apples, 0 otherwise
coop = dummy variable: 1 if co-operative store, 0 otherwise
The expected signs on the coefficients of the variables are as follows:
var(+), num(-), shop(-), size(-), org(+), mark(-), v1(0), v2(0), v3(0), v4(0), v5(0), v6(0), coop(0)
The variety of items should increase the overall prices for several reasons. First, variety of offerings can take the place of low prices as an inducement to shop at the store or market. Second, the extra variety will tend to come from specialty items which are purchased less frequently and which may also (but not necessarily) be more fragile. These characteristics will lead to higher produce losses, which must be compensated for by higher prices. The number of competitors should lower prices as they increase the supply and compete for business. The number of shoppers could raise prices by increasing demand, but will more likely allow for greater sales per sales area and per hour worked, reducing the amount of fixed cost (such as store rent) that must be covered by each individual sale. Size should have a negative effect on prices because large size allows for bulk purchases and greater turnover for stores, or increased mechanization of farm operations and efficient use of management abilities for farmers. Organic produce is generally more expensive than conventional produce, and results from this study are not expected to be different. The variable for cooperatives is not expected to have an effect on prices, for reasons discussed below. Finally, the type of vegetable (v1 - v6) should not affect the overall result.
Early models with this data were heteroskedastic. This occurred either if the data was indexed with a sample mean taken across all locales, or if a single regression was attempted for all the data. This heteroskedasticity was determined by a chi-squared test. Despite the lack of a uniform variance, a Chow test was performed on the data; not surprisingly, it showed the data points breaking down by locale. Part of the problem was that price differentials between the farmers' markets and the stores did not correlate with variables collected to describe the different locales, i.e. population sizes and income levels. Tastes and preferences marginally related to these simple statistics or variables unmeasured here greatly affect pricing strategies. I chose therefore to perform a separate regression for each locale, with prices indexed as indicated above.
Furthermore, several variables were eventually omitted from the final models. The dummy variables for the individual vegetables (v1 - v6) had to be dropped in order for the several estimations with the reduced number of observations to solve. The variable num was found to co-vary significantly with everything but income and store size. It was therefore dropped because it over-specified the model. This action had almost no affect on the goodness of fit, but did improve the t-ratios of other variables. An additional reason for it being dropped was that its specification was problematic. For a store, the variable was defined as number of competing stores; for a market, it was number of stalls selling more than just a few items. The intention was to capture the effects of competition, for it has been shown that concentration of the retail grocery industry in an area leads to higher prices (Marion, et al., 1979). However, competition between stalls and that between stores is not directly comparable. A variable to capture competition would need to be more complex.
Two other omitted variables were shop and coop. Shop had significant covariances with nearly everything, so it was dropped as a redundant variable. The unfortunate consequence of removing shop is that the utility derived from attending the farmers' market because of its status as a popular community event is not separated from the utility derived from purchasing fresh produce direct from the grower. The variable coop was abandoned early. First, there were only two co-operatives in the data set, out of fifteen total stores, so the number of observations was small. Second, casual observation by the author and others has failed to note much difference between co-ops' prices and those of grocery stores. Since no significance was anticipated, when early estimations ascribed to coop a low t-score as well as a low coefficient, it was removed from further regressions. The removal had almost no effect on the estimation.
The results of the regressions for the individual locales demonstrate that the structure of the model obviously changes by area. This confirms the problems of correlating the price differentials to the statistics used to describe the locales, i.e. population and income level. Variables which are highly significant in one locale cease being significant in another. Every variable but org has both positive and negative signs, and that variable varies in magnitude. Even the most important variable for the purposes of this paper, mark, changes sign, size, and significance. These results demonstrate the complexity of pricing strategies, in particular those of farmers' markets.
That said, it is important to note that farmers' market prices are generally lower. For four of the five locales, the coefficient for the variable mark is negative and the absolute value of the t-ratio is greater than unity; for two of them, the t-ratio is significant at the 5% level. Only one does not show this tendency, and that one indicates parity between market and store prices. The results, in fact, confirm the impressions that this author has formed from shopping and working at the different markets. At Marin, the market is held in the parking lot of the Civic Center designed by Frank Lloyd Wright. It dedicates a lot of space to non-farmer vendors selling artsy hats, specialty vinegars and dressings, and other craft items. Many of the shoppers are economically comfortable. They enjoy the bazaar atmosphere or European flavor of a market, and running into each other there seems to have a certain cachet. Consequently, farmers are not as hard-pressed to offer bargain prices to assure a clientele, because shoppers derive utility from aspects of the market other than the low prices. The contrary case is Sacramento, where the Sunday market takes place in a parking lot under the freeway. Nearly the entire space is dedicated to farmers, the exceptions being the fish vendors and the two or three local bakeries which have stands. Shoppers tend to be less well-off economically, and many are from the ethnic communities of the city. They are concerned with freshness and price, as well as with finding produce items that grocery stores might not carry. It is therefore not surprising that the produce was much cheaper at the market, and that the t-ratios showed very high significance. In this vein, it is interesting to note that the Sunday Sacramento market does not have a single certified organic farmer.
Perhaps an economic model of farmers' market prices needs to include more variables to be generally applicable. What is more likely is that the effect of the market on prices depends on the character of the locale, on the weight it places on direct marketing, the event of the market, and price savings. As for the central question of the paper, whether prices at farmers' markets are lower, it seems that they by and large are lower, though by how much will vary according to area. An interesting direction for more study would be to amass new time-series data to compare to the set from Sommer, et al. (1980), to try to uncover changes which have occurred in the interim. Furthermore, attempts should be made to identify sociological variables affecting farmers' market prices, especially shoppers' perceptions and their reasons for participation. These could help clarify the complex good which people are "purchasing" when they shop at a farmers' market. This information can help community and city planners articulate their goals when starting a market (is it to be a source of low- cost food? a community center? both?) and deciding how to design it, where to hold it, and what vendors to allow participate.
Clifford M. Guy and L. G. O'Brien. "Measurement of Grocery Prices: Some Methodological Considerations and Empirical Results." Journal of Consumer Studies and Home Economics 7, no. 3 (Sept. 1983): 213-227.
D. L. Maley. "Retail Grocery Price Variation in Saskatchewan." Canadian Farm Economics 11, no. 4 (August 1976): 23-29.
B. W. Marion, W. F. Mueller, R. W. Cotterill, and F. E. Geithman. "The Price and Profit Performance of Leading Food Chains." American Journal of Agricultural Economics. 61, no. 3 (August 1979): 420-433.
Robert Sommer, Margot Stumpf, and Henry Bennett. "Quality of Farmers' Market Produce: Flavor and Pesticide Residues." Journal of Consumer Affairs. 16, no. 1 (Summer 1982): 130-136. In a double-blind taste test in Davis, CA, the researchers found that for four of nine produce items, people preferred those bought at the farmers' market over those bought at local stores. For the remaining five items, there was no significant preference.
--------- , Margaret Wing, and Susan Aitkens. "Price Savings to Consumers at Farmers' Markets." Journal of Consumer Affairs. 14, no. 2 (Winter 1980): 452-462.
William H. (Bill) Blake III first noticed market pricing differences when working on an organic farm. A Virginia native, he earned his M.S. at the University of California, Davis in International Agricultural Development. He is currently involved in several projects for UC Cooperative Extension (UCCE) and UC Davis, and is the main author of Making the Connection, a UCCE handbook for Community Supported Agriculture. UCCE Placer County, Dewitt Center, 11477 E Avenue, Auburn, California 95816
WK: 916/889-7385; E-mail: Bill Blake <ceplacer@ucdavis.edu> ; Home phone: 916-451-4852 (Sacramento)