Dublin pphg 140807-soudek-skuhrovec

Public Procurement of Homogeneous Goods: an Empirical Study

Jan Soudek & Jiří Skuhrovec, Charles University, 8/2014

[email protected] , [email protected]

Outline

• Introduction

• Motivation & literature review

• Hypotheses & model

• Data description

• Results & discussion

2

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

INTRODUCTION: SIZE OF THE MARKET & INSTITUTIONAL FRAMEWORK

3

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

Public procurement to GDP (2011)

Source: OECD: Government at glance 2013

In case of CZ, it is app. 16% of GDP (excluding SOE and utilities, then it would be around 25%)

4

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

0%

5%

10%

15%

20%

25%

Institutional framework

• 3 basic procedures of tendering

• Open – The most transparent, everyone can apply for that procurement

• Restricted – Procurer chooses potential suppliers

• Negotiated – Procurer can ask just one firm to supply it‘s demands.

• 2 basic formats

• Sealed bid auction - send an envelope

• Electronic auction - a tool allowing repetitively adjust the bid

5

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

Procurement (auction) theory

• Procurement procedure: • standardized, trading mechanism

• subject of trade is known ex ante

• Thus auction theory is applicable

Evaluation process

Bid 3

Bid 1 Bid 2

Winner

6

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

MOTIVATION & LITERATURE REVIEW 7

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

Motivation

• What determines the final price in public procurement? Market price? Sure, but what about:

• Chosen procedure

• Competition

• Auction format

• Our goal: show where institutional and procedural characteristics matter in the public procurement.

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

8

27

.8.2

01

4

Previous research on matter • Domberger et al (1995): „The Determinants of Price and Quality in Competitively

Tendered Contracts “ EJ 1995 • Model:

• Conclusion:

• While competition reduced price significantly, quality of service was maintained or even enhanced.

• The evidence does not support the hypothesis that efficiency gains are traded off for lower quality.

• Bandeira et al (2008): „Active and passive waste in Government spending: Evidence from a policy experiment“ AER 2009 • Model:

• Conclussion:

• Some public bodies pay systematically more than others for observationally equivalent goods and such price differences are sizeable;

• Differences are correlated with governance structure: the central administration pays at least 22% more than semi-autonomous agencies (local government is at an intermediate level)

9

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

Theoretical background - Auction theory says:

• Assuming standard auction with interdependent signals:

• With # of bidders the price (in average) goes down (Bower (1993), Bulow&Klemperer (1996))

• Open procedure = competition leads to better results than negotiation (Bulow &Klemperer (1996))

• Electronic auction = English auction results are at least as good as first price sealed bid auction - regular „envelope“ procedure (Milgrom&Weber (1982))

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

10

27

.8.2

01

4

Our research hypotheses

H1: The final unit price of the procurement is affected by the type of procurement procedure.

H2: The final unit price of the procurement is a decreasing function of an number of bidders.

H3: The usage of electronic auction is decreasing the final unit price of the procurement.

H4: The final unit price of the procurement is affected by the type of the contracting authority.

27

.8.2

01

4

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

11

Model 1. In order get comparable prices, only PP of homogeneous goods is

analyzed – electricity, gas

We can get the unit prices and compare them with actual market price (1 year forward contract)

Moreover, we compare with estimated value per unit – which captures for potential heterogeneity of tendered goods and for procurers willingness to pay

2. We use a natural logarithmic form of final unit price as a dependent variable to capture relative effects

3. To control for possible „wasteful behavior“, we use zIndex; a composite index that rates procurers according to quality and transparency of all their procurement competitions over given period of time.

𝑙𝑜𝑔 𝑓𝑖𝑛𝑎𝑙 𝑢𝑛𝑖𝑡 𝑝𝑟𝑖𝑐𝑒 = 𝛼 + 𝛽1 𝑙𝑜𝑔 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑢𝑛𝑖𝑡 𝑝𝑟𝑖𝑐𝑒 +

𝛽2 𝑙𝑜𝑔 𝑚𝑎𝑟𝑘𝑒𝑡 𝑢𝑛𝑖𝑡 𝑝𝑟𝑖𝑐𝑒 + 𝛽3𝐷𝑈𝑀𝑀𝑌 𝑝𝑟𝑜𝑐𝑒𝑑𝑢𝑟𝑒 + 𝛽4𝐷𝑈𝑀𝑀𝑌 𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑛𝑖𝑐 𝑎𝑢𝑐𝑡𝑖𝑜𝑛 +

𝛽5𝐷𝑈𝑀𝑀𝑌 𝑎𝑢𝑡ℎ𝑜𝑟𝑖𝑡𝑦 + 𝛽6𝑏𝑖𝑑𝑑𝑒𝑟𝑠 + 𝛽7𝑧𝐼𝑛𝑑𝑒𝑥 + 𝛽8𝑡𝑖𝑚𝑒 + 𝜖

27

.8.2

01

4

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

12

DATA DESCRIPTION 13

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

Data description

• Source isvzus.cz (Czech TED)

• Tenders above limit of 2mil CZK (80k €)

• Years 2008 – 2013

• 220 Electricity & 57 Gas competitions

• Total worth 10bn CZK (400 mil. €)

• Metric of market price is one year forward price (base CAL) from Prague Exchange

27

.8.2

01

4

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

14

Electricity auctions in time

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

15

27

.8.2

01

4

Institutional characteristics

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

16

• 80 % auctions using open procedure

• 49 % auctions using e-auction

• Most frequent contracting authorities are SOE and regional authorities (Municipalities, regional offices etc.)

27

.8.2

01

4

RESULTS & DISCUSSION 17

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

Results

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

18

27

.8.2

01

4

Dependent variable:

log(final price/kWh) Electricity Gas

Explanatory variable OLS ß Robust SE OLS ß Robust SE

log (estimated price/kWh) 0.43 (0.09) *** 0.41 (0.11)***

log (market price/kWh) 1.03 (0.19) *** 0.15 (0.09)*

Open procedure -0.07 (0.04)* 0.08 (0.06)

electronic auction -0.12 (0.03)*** -0.15 (0.06)**

number of bidders -0.015 (0.008)* -0.03 (0.01)*

Constant -0.10 (0.07) -0.11 (0.08)

R-squared 0.54 0.52

Table 2: Electricity & gas procurement results, note: dropped dummy is negotiated procedure

Results - regression with insignificant variables

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

19

27

.8.2

01

4

Dependent variable:

log(final price/kWh)

Electricity Gas

Explanatory variable OLS ß Robust SE OLS ß Robust SE

log (estimated price/kWh) 0.41 0.1*** 0.49 0.11***

log (market price/kWh) 1.01 0.23*** 0.1 0.12

Open procedure -0.08 0.04* -0.03 0.018

electronic auction -0.16 0.04*** -0.13 0.07*

number of bidders -0.016 0.008* 0.09 0.07

zIndex -0.14 0.94 0.27 1.01

Time 0.00003 0.00004 0.00009 0.00010

State owned enterprises -0.016 0.07 0.28 0.17

Public bodies 0.002 0.09 0.19 0.19

Regional auth. 0.06 0.08 0.2 0.18

Constant 0.13 0.17 -0.48 0.25*

R-squared 0.55 0.59

Table A-1: Electricity & gas procurement results – including Procurers, note: dropped dummies are

negotiated procedure and regional authorities

Discussion

• Final price rises by 0.4 % with 1 % rise in estimated price. Signalling?

• Open procedure reduces cost of electricity by 7 % in average, for gas it is insignificant.

• E-auction reduces cost of electricity by 12 % in average (cumulative with Open procedure!), gas by 15 %.

• Additional bidder causes in average a 1 % fall in the final price of the electricity and 1.5 % fall in final price of gas

• We did not find any statistically significant difference in final prices with respect to the various types of contracting authority.

27

.8.2

01

4

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

20

Policy implications

• Buyer should identify the estimated price well, not overshoot. It signals: • price expectation

• willingness to pay.

• Electronic auction shows to be most efficient. Use it. • 12 % average price reduction

• Low additional procedure costs (10.000 CZK = 400 €)

27

.8.2

01

4

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s

21

Questions? Thanks for attention!

22

Pu

blic

Pro

cure

men

t o

f H

om

oge

neo

us

Go

od

s 2

7.8

.20

14

All details & sources are here: http://ies.fsv.cuni.cz/sci/publication/show/id/4833/lang/en Contact: Jan Soudek, Jiri Skuhrovec Charles University [email protected] , [email protected]