Author: Publisher:
ISBN:
Category :
Languages : en
Pages : 0
Book Description
I study the effect of lump sum contracts on bidding behavior and procurement costs in highway construction. For highway construction projects we usually observe ex-post changes in design due to unforeseen site condition. The Florida Department of Transportation (FDOT) offers two different contract types regarding compensations for ex-post quantity changes of work item. Under the standard contract, the department provides a list of work items and their quantities. If there is a quantity change, the contractor's unit price bid on that work item is used to make pay adjustment. On the other hand, under the lump sum contract, bidders are allowed to design the project themselves and no pay adjustments are made for ex-post quantity changes. In Chapter 1, I propose a theoretical model to describe bidding behavior under the lump sum. They are willing to inflate bids to self-insure against ex-post changes, while they should make a cost-efficient design to win the contract. Compared to the standard, lump sum contracts may increase or decrease bids, but they reduce the expected payments in many cases. In Chapter 2, I estimate the effect of lump sum contracts on (winning) bids and procurement costs. I assume that the outcome variables are independent of the assignment conditional on observables. And I use regressions and propensity score methods to estimate the effects. Specifically, I use a trimmed sample for regressions in which some seemingly extreme observations are dropped based on the propensity scores. With miscellaneous projects, lump sum does not increase the outcome variables. And I allow the assignment decision to be endogenous and use the treatment effect model to quantify the impact. In particular, I use district and task dummies as instrumental variables. It turns out that the outcome variables do not increase under the lump sum. And I test for unobserved project heterogeneity conditional on all the covariates by exploiting multiple bids within an auction. Using the FDOT data, the correlation coefficient for the trimmed sample is close to zero, which implies no serious unobserved project heterogeneity. Thus, the regression results with the trimmed sample are supported.
