What Are the Priorities of Public Buyers? Evidence from a Conjoint Experiment with Bureaucrats in Finland, Germany, and Italy

A conjoint survey experiment (choice experiment) as detailed below.

2020-09-09

2020-12-31

The primary outcome considered is the binary choice regarding which of the two respective decision scenarios concerning tender outcomes is desirable in the eyes of the procurement bureaucrat. Procurement outcomes are defined by six experimental features which affect the utility of these outcomes. These six experimental attributes to procurement outcomes are the price of the purchase as stated in the winning bid, quality as promised in the winning bid, the number of submitted bids, whether judicial complaints were filed to challenge the tender, geographical location of the winning firm, as well as the procurement official’s previous experience with the winning bidder.

The study involves a conjoint survey experiment that is carried out among a non-probability sample of procurement employees in Finland, Germany, and Italy. We reach out to procurement officials through collaboration partners in each country (Finnish Competition and Consumer Authority, Deutsche Vergabenetzwerk in Germany, and Telemat in Italy) via e-mail and send invitations to participate in our online survey, embedding the conjoint scenarios, using individualized web-links for each respondent.

First, all respondents are presented with an introductory text identical across experimental conditions. Second, all survey respondents first answer (optional) background questions about their socio-demographic, job- and office-characteristics. For example, we ask about their education, job experience, and individual career concerns and status. Moreover, we ask about their office-characteristics (e.g., office size or type) as well as the typical tendering procedure (e.g., open vs. restricted) and the awarding rule officials typically work with (lowest price vs. scoring rule), and typical tendering outcomes relating for example to the level competition and occurrence of litigation. Third, all respondents are presented with an identical introductory text to the conjoint part of the study. Fourth, all respondents face a series of decisions between two fictitious profiles of tender outcomes. For each scenario, the respondents have to decide which tender profile they would prefer to have realized. Profiles differ with respect to several attributes that are randomized by the researchers. Moreover, the order of the attributes is randomized across respondents but is fixed at the respondent level.

Specifically, we ask the following outcome question after each decision scenario:
“Please look at the following pair of hypothetical tender outcome scenarios carefully and make a decision which you would like more. Which tender outcome scenario do you prefer?”

We employ the following attributes and respective realizations in each of the 6 hypothetical decision scenarios:

The price as stated in the bid from the winning firm is:
 much lower than I expected
 a bit lower than I expected
 what I expected
 a bit higher than I expected
 much higher than I expected

The quality of the bought good or service as promised in the bid from the winning firm is:
 as I expected
 a bit better than I expected
 much better than I expected

As a result, the tender received …
 ... 1 bids.
 ... 2 bids.
 ... 4 bids.
 ... 8 bids.

The selected winner is …
 … is a firm that was unknown to me through previous tenders.
 … is a firm I already knew from previous tenders and trusted.
 … is a firm that I already had a bad experience with.

The winning firm is:
 A local bidder from your region
 A non-local bidder that does not come from your region

After awarding the contract, ... legal complaint has been filed against the tender.
 … no … [weighted probabilities of 90%]
 … yes … [weighted probabilities of 10%]

Respondents then select between the depicted tender outcomes of profile A or B.
Finally, after the conjoint analysis, we ask respondents which of these attributes are most and least important (i) for achieving a desirable tender outcome, (ii) in their daily work, and (iii) in improving their career prospects. We randomize the order of these follow-up questions to study the mechanisms of stated choices and to distinguish between choices, work practices, and career concerns of bureaucrats.

Uniform randomization of choice scenario from the set of possible choice scenarios based on a professional survey software.

Randomization is at the individual choice scenario level

No

Number of individuals in our surveys. We field the survey to approx. 1,300 procurement employees in Finland, 2,000 in Germany, and 18,000 in Italy. The exact number of survey invitations is unknown at this point, because the contact information data bases of collaboration partners are constantly updated. We do not know ex-ante the response rate but expect about 2,000 completed surveys embedding our experiment. We expect the response rate to be the highest in Finland as there we use endorsement letters to boost the response rate.

12,000 choices across 2,000 individuals.

Uniform randomization of 720 possible pre-determined choice scenarios across 12,000 choices (6 choices per expected individual response).