How to Achieve Efficiency in Public Procurement Auctions

In the wake of the recent crisis many countries face problems caused by budget deficits. To be able to lower their budget deficits these governments should conduct their expenditures at the lowest possible prices. To achieve this objective many countries use auctions to administer government procurements. Turkey is one of the countries where government procurement is conducted mainly using first-price auction methodology. In this study, we make use of a unique data set provided by the Public Procurement Authority (PPA) of Turkey which comprises detailed information about all GP auctions for the years 2004-2010, 472560 first-price auctions. Using this data set, we empirically investigate the the optimal competitive environment for lowest procurement costs. First, we analyze the effect of number of bidders on the procurement price (winning bid). Then we examine the main research question of the paper and investigate the optimal number of bidders for different types of products which renders lowest procurement costs. In order to examine the research questions raised in the introduction section we conduct the following analysis. First, we run two sets of regressions in order to separately investigate the effects of our explanatory variables on the bidders' decision to enter an auction and how the auction prices are determined. Following Bajari and Hortacsu (2003), we use a negative binomial regression model to analyze bidders' entry decision and how auction specifications affect the number of participants. Then we conduct the auction price determination regression to analyze the determinants of auction prices and the effect of number of bidders on auction prices. We take into account the possible endogeneity. Endogeneity problem might affect the empirical results since unobserved variables correlated both with the number of bidders and with the auction price might exist. We implement the GMM methodology to control for endogenous regressors. Finally, we search for the optimal number of bidders for each product type by comparing the mean of dependent variables for different number of bidders. Estimation procedures and the results are discussed in detail in the following sections. Using this unique data set, we first analyze the determinants of number of bidders. Then, we focus on the effect of competitive environment on the procurement costs by examining the effect of various explanatory variables on the difference between the contract price and the estimated cost of auctions. Finally, we analyze the optimal number of bidders for different product types. We take into account the endogeneity of the number of bidders while conducting these analysis. We have two major results. First, we show that the number of bidders significantly and negatively affects the procurement price. Thus, existence of a more competitive environment significantly decreases government procurement costs in Turkey. Second, the optimal number of bidders to take the full advantage of competition differs among auctions for different types of products. At least nine bidders are needed for services; six bidders are required for the goods and construction auctions to be able to achieve the lowest procurement price possible. The findings of our study are in line with the standard theoretical predictions of the IPVP auction models and confirm some of the key empirical findings of previous studies like Iimi (2006).