Economy and Food Security

Introduction and Objectives: Fresh agricultural products such as tomatoes are highly perishable, with rapid post-harvest quality deterioration and sensitivity to environmental conditions. According to the FAO (2024), approximately 14% of global food production is lost post-harvest, with substantial waste at the retail level. Tomatoes, a staple in Iranian households, are extremely sensitive to respiration, temperature, and tissue softening. Dynamic pricing through daily price adjustments is a key strategy for inventory management and waste reduction. While extensive literature exists on the "depth of discounts," few studies have addressed the fundamental question of what determines how many times a seller changes the price during a day. This gap motivates the present research. The main objective is to identify and analyze the determinants of the number of daily price levels (the frequency of tomato price changes) in three fruit and vegetable stores in Gorgan, Iran. The core novelty is introducing the "number of price levels" not just as a managerial tool, but as a multi-factor behavioral phenomenon modeled using daily micro-data.
Materials and Methods:  This research is applied in terms of purpose and quantitative in methodology, based on daily panel data. Required data were collected during the year 1402 (2023 AD) from three licensed fruit and vegetable stores in Gorgan, which were randomly selected from active stores. The dependent variable, "number of daily price levels," was defined with three nominal levels including 1 (one fixed price throughout the day), 2 (two distinct prices, e.g., morning and evening), and 3 (three or more prices during the day). Independent variables included the initial price of tomatoes at the beginning of the day (thousand Tomans), store identity (Store 2 compared to Store 3 as the base group), weekly day patterns (Monday-Tuesday compared to Friday, and Wednesday-Thursday compared to Friday), and the natural logarithm of the ratio of leftover product at the end of the day to the initial purchase quantity. Since logarithmic transformation requires positive values, 42 observations with zero or negative ratios (usually due to zero waste or data recording errors) were excluded from the dataset, and the final model was estimated on 150 store-day observations. Although the dependent variable has a natural order, due to the unequal intervals between levels and the differing decision-making mechanisms for moving from one to two levels versus from two to three levels, a multinomial logit model with emphasis on the Independence of Irrelevant Alternatives (IIA) assumption was employed instead of ordered models. The base level in this model was set to one price level per day. Parameter estimation was performed using Stata software via the Maximum Likelihood method. To check the robustness of the results, extra control variables such as temperature, sales-to-purchase ratio, holidays, and lagged waste variable were tested in initial estimations but were excluded from the final model due to statistical non-significance or convergence issues.
Results: Descriptive statistics revealed that the dominant pricing pattern in the Gorgan tomato market is the use of two price levels during the day, accounting for over 75 percent of observations (113 out of 150 observations). In contrast, days with one price level and three price levels accounted for 12 percent (18 observations) and 12.67 percent (19 observations) of total observations, respectively. Furthermore, it was observed that as the number of price levels increased from one to three, the average ratio of leftover product at the end of the day decreased from approximately 18 percent to 11 percent, indicating the relative efficiency of more dynamic pricing in inventory management. The estimation results of the multinomial logit model showed that the model is overall well-fitted (LR chi2(10) = 82.50, Prob > chi2 = 0.0000), with a Pseudo R² value of 38.35 percent, indicating that the explanatory variables explain a substantial portion of the variation in the dependent variable. The main findings are as follows: (1) Initial price has a negative and significant effect on the probability of using both two-price (0.304-) and three-price (0.210-) levels, meaning that each one-thousand-Toman increase in the initial price reduces the seller's tendency to change prices. (2) Store 2, compared to Store 3 (the base), shows a significantly greater tendency to use multi-level pricing (coefficients of 2.018 for level two and 3.231 for level three), reflecting behavioral and managerial differences among stores. (3) Wednesdays and Thursdays (close to the weekend holiday) significantly increase the probability of price changes, consistent with increased demand patterns at the end of the week. (4) Contrary to theoretical expectations, a higher ratio of leftover product at the end of the day has a negative effect on the number of price levels, which may be due to retailers following fixed, pre-determined pricing rules. (5) The effect of Mondays and Tuesdays was only marginally significant for the two-price level.
Conclusion: Retailers' decisions to change prices frequently are a complex process, simultaneously influenced by cost signals (initial price, potential waste), time signals (late-week days), and store-specific institutional characteristics. These findings align with classical dynamic pricing and perishable inventory literature but extend it by focusing on the "frequency of change" as the dependent variable. From a managerial perspective, stores facing surplus inventory are advised to increase price change frequency (especially on Wednesdays and Thursdays) rather than reducing profit margins early. The study is limited by its small sample size (only three stores) and lack of qualitative variables (product grading, competitor prices). Future research should expand samples across cities and employ quasi-experimental methods to establish causality between price levels and waste reduction.