Mihály Fazekas - István János Tóth

Developing and Implementing

Data Proxies for Detecting Corruption

Mihály Fazekas*- István János Tóth+

*: University of Cambridge and Corruption Research Center Budapest, [email protected]

+: Hungarian Academy of Sciences and Corruption Research Center Budapest, [email protected]

2014.12.04. 1

Salzburg Data Science Symposium,

Salzburg, 20 November 2014

This project is co-funded by the

Seventh Framework Programme for

Research and Technological

Development of the European Union

mailto:[email protected]






"When you can measure what you are speaking

about, and express it in numbers, you know

something about it; but when you cannot express

it in numbers, your knowledge is of a meagre and

unsatisfactory kind; it may be the beginning of

knowledge, but you have scarcely, in your

thoughts, advanced to the stage of science,

whatever the matter may be."

Lord Kelvin

Idea vera debet cum suo ideato convenire

[Spinoza: Ethica; Axiomata VI.]

Staff & Support

Staff:

Ágnes Czibik economist

Mihály Fazekas Ph.D., University of Cambridge; head of research CRCB

Gyula Fóra ELTE, student

Júlia Orbán economist

Bálint Szalai journalist

Bence Tóth economist

István János Tóth Ph.D, senior research fellow, IE HAS; director CRCB

Experts:

Zoltán Kelemen lawyer

Jenő Gyenese computer programmer

Zoltán Nagy ecomomist (former head of Hungarian Competition Authority)

Sándor Rácz computer programmer

Zoltán Siposs journalist

Tamás Uhrin computer programmer

Data cleaning: Balázs Agárdy, Mónika Bak, Mária Balla, Eszter Bíró, Linda Borbély, Eszter Csizmás, Kinga Csizmás, Olivér Csukás, Eszter Farkas, Annamária Ferenczi, Katalin Gajdos, János Gáspár, Diána Groszmann, Tamás Heizer, Erik Herbák, Erzsébet Hoffman, Eszter Kádár, Emese Koplányi, Gabriella Korom, Balázs Kovács, Dávid Levendel, Máté Magyar, Anna Markó, Eszter Milibák, Ákos Nagy, Laura Pacsa, Ilona Pallagi, Tibor Pallagi, Ágnes Parkot, Erika Révész, Flóra Samu, Levente Séd, Eszter Simon, Dóra Somogyi, Nóra Staub, Adrienn Süli, Tímea Szabó, Adrienn Szigili, Győző Teplán, Petra Vancsura, Attila Varga, Kinga Varga

Support:

voluntary work, EU FP7, TAMOP, Hungarian Competitive Authority

2014.12.04. 4

Overview

• Motivation

• Corruption at a Glance

• Measurement: old & new method

• Database building

• Proposed corruption risk indicators

• Illustrative results

• Further research

2014.12.04. 5

motivation corrupion measurement data buiding indicators illustrations plans

Motivation

motivation corrupion measurement data indicators illustrations plans

Policy problems

• How much public money is lost due to

corruption in public procurement?

• Which tender should be investigated by

the authorities?

• What works for reducing corruption risks in

public procurement?

2014.12.04. 7


Why public procurement?

1. A lot of money involved

2. Crucial role in development (e.g. capital

accumulation)

3. Indicates the broader quality of state

institutions

2014.12.04. 8


Why public procurement?

4. Very corrupt

2014.12.04. 9


Corruption at a Glance



• „The misuse of public/private [entrusted]

power power for private gain”

• Actors: principal - agent - client PA model from the micro-economics

(e.g.: government – gov. officials – comp. manager)

• Petty corruption vs. Grand corruption

2014.12.04. 11



• Where could occur?

– Issuing licences

– Inspection (police, tax authorities, e.t.c.)

– Regulation, Legistlation

– Public Procurement

2014.12.04. 12



• Main types:

– bribery

– extortion

– emblezzement

– fraud

2014.12.04. 13


Definition of instutionalised grand corruption

• Specific definition (just like measurement)

• Institutionalised grand corruption in public procurement

institutionalised grand corruption in public procurement refers to the regular particularistic allocation and performance of public procurement contracts by bending universalistic rules and principles of good public procurement in order to benefit a group of individuals while denying access to all others.

2014.12.04. 14


Measurement of Corruption


Old methods

• Available indicators are either biased or too idiosyncratic

– Perception-based survey instruments measure PERCEPTIONS (e.g. TI CPI)

– Experience-based survey instruments suffer from conformity bias and lack of access

– Audits and case studies measure narrow phenomena

Need for new indicators

2014.12.04. 16


New methods for measurement of Corruption

• Perception indicators are not good enough

• New paradigm of measurement

– harnessing BIG DATA,

– built on thorough understanding of context,

– understanding mechanisms

– and

– ‚open-ended’

2014.12.04. 17


Our philosophy for measurement

• Corruption is a hidden (exremly latent)

phenomenon, it is impossibe to observe at

actors’ level

• Instead to observe corruption we observe

the risk of corruption

• Informations from micro (actors’ ) level

2014.12.04. 18


Corruption like a Black Hole

2014.12.04. 19


New approach for measurement

• Our Indicators’ characteristics:

– Specific

– Real-time

– ‘Objective’/hard

– Micro-level

– Aggregatable + comparative

2014.12.04. 20


Building a linked database

• Public procurement data

• Company financial and registry data

• Company ownership and management data

• Political officeholder data

• Treasury accounts of public organisations

• Arbitration court judgements

2014.12.04. 21


PP data - State of the art

• Global standardization effort: www.open-contracting.org/

• TED data: DG Markt, http://ted.openspending.org/, etc.

• National datasets:

– Developed economis: CA, IT, US

– Transition economies: CZ, HR, HU, RO, SK, SL

– Developing countries: GEO: http://tendermonitor.ge/en

• Development agencies

– World Bank

– UNDP

2014.12.04. 22


http://www.open-contracting.org/




http://ted.openspending.org/


http://tendermonitor.ge/en

Initial data and database building


The Worst and the Ugly • The Hungarian authorities do not have a

comprehensive database about PP (from 1998- )

• Majority of information is string, not numeric

• Only individual cases => case studies

• Lack of measurement and => lack of reliable analysis

• Weak knowledge about causes and mechanisms of corruption in PP

2014.12.04. 24



Hungary: aggregate data [pdf]

http://www.kozbeszerzes.hu/static/uploaded/document/Sajt%C3%B3anyag_2011%20%C3%

A9v%202012%20%20els%C5%91%20negyed%C3%A9v_v%C3%A9gleges.pdf


http://www.kozbeszerzes.hu/static/uploaded/document/Sajt%C3%B3anyag_2011 %C3%A9v 2012 els%C5%91 negyed%C3%A9v_v%C3%A9gleges.pdf

http://www.kozbeszerzes.hu/static/uploaded/document/Sajt%C3%B3anyag_2011 %C3%A9v 2012 els%C5%91 negyed%C3%A9v_v%C3%A9gleges.pdf

Hungary: individual data [txt, html, php]

http://www.kozbeszerzes.hu/ertesito/megtekint/portal_15125_2012/


http://www.kozbeszerzes.hu/ertesito/megtekint/portal_15125_2012/

First Period - 1998-2004: txt


Second Period - 2005-2012: html


Third Period - 2013 - : html, xml


An ideal case: Chile

http://www.chilecompra.cl/


http://www.chilecompra.cl/

MaKAB: 1998-2014


Our tasks • Database architecture (MySQL)

• Dowloading text, html, xml data form HPP webpage

190.000 observations and cca. 300 variables

• data cleaning (Python, Java, PHP)

(e.g. „Budapist”, „Budipest”, „Budpest” to „Budapest”)

• Data corrections: e.g. values in EUR, HUF, USD => HUF

with and without VAT => net values

• And so on…

2014.12.04. 33


Our tasks • Tests and data improvement (imput data from external

sources: e.g. company data)

• Indicator building

• Indicator tests, validation

• Analysis (STATA, SPSS, R)

2014.12.04. 34


Building a linked database

• Public procurement data

• Company financial and registry data

• Company ownership and management data

• Political officeholder data

• Treasury accounts of public organisations

• Arbitration court judgements

2014.12.04. 35


PP data - State of the art

• Global standardization effort: www.open-contracting.org/

• TED data: DG Markt, http://ted.openspending.org/, etc.

• National datasets:

– Developed economies: CA, IT, US

– Transition economies: CZ, HR, HU, RO, SK, SL

– Developing countries: GEO: http://tendermonitor.ge/en

• Development agencies

– World Bank

– UNDP

2014.12.04. 36






http://tendermonitor.ge/en

Proposed indicators


Blueprint for measuring institutionalised

grand corruption in PP

1. Corruption Risk Index (CRI): generation and allocation of rents [intervall]

2. Political Influence Indicator (PII): political influence on companies’ market success [intervall]

3. Political Control Indicator (PCI): direct political control of contractors [0,1]

2014.12.04. 38


CRI

• A composit indicator (information from elementary indicators)

• Information from the PP documents (tender docs, assessment,

delivery )

• Daily data

• Micro (from micro to macro level)

• Initiative step: CRI at contract level

• And:

CRI at issuer level and

CRI at winner, competitor level

CRI at country level


CRI

• Number of bids (if 1 => high corruption risk)

• Call for tender not published in official journal (if yes => high CR)

• Procedure type (if not open => high CR)

• Length of eligibility criteria (if long => high CR)

• Exceptionally short submission period (if yes => high CR)

• Relative price of tender documentation (if high => high CR)

• Call for tenders modification (if yes => high CR)

• Weight of non-price evaluation criteria (if high => high CR)

• Annulled procedure re-launched subsequently (if yes, high CR)

• Length of decision period (if short => high CR)

• Contract modification (if yes => high CR)

• Contract value increase (if yes => high CR)

• Winner's market share (if high => high CR)


Share of PP with one bidder by month,

in Hungary, 2009.01-2014.10, N = 108.287

An elementary indicator: number of bids motivation corrupion measurement data indicators illustrations plans

Corruption Risk Index (CRI) • Probability of institutionalised grand corruption to occur

0 ≤ CRIt ≤ 1

where 0=minimal corruption risk; 1=maximal observed corruption risk

• Composite indicator of elementary risk (CI) indicators

CRIt = Σj wj * CIj t

• Elementary risk indicators are combined to reflect a corrupt rent extraction logic

2014.12.04. 42


Components of CRI

2014.12.04. 43

X = available

* = potentially available

CZ HU SK TED

Single bidder contract (valid/received) 1=1 bid received x x x x

Call for tenders not published in official journal 1=NO call for tender published in official journal x x x x

Procedure type open, invitation, etc. x x x x

Relative length of eligibility criteria Number of characters relative to market average x x * *

Length of submission period Number of days between the publication of call for tenders and the submission deadline x x x x

Relative price of tender documentation price of tender documentation DIVIDED BY contract value x

Call for tenders modification 1=call for tender modified during submission period x x * *

Exclusion of all but one bid 1=all but one bid excluded x

Weight of non-price evaluation criteria Sum of weights for evaluation criteria which are NOT related to prices x x x x

Annulled procedure re-launched subsequently 1=contract awarded in procedure annulled, but re-launched x

Length of decision period number of days between submission deadline and announcing contract award x x x x

Contract modification 1=contract modified during delivery * x

Contract lengthening/value increase relative contract extension (days) or relative contract price increase * x

overall winner contract share 12-month total contract value of winner / 12-month total awarded contract value (by issuer) x x x *

9 14 7 6

11 14 9 9

Number of availab le components

Number of potentially availab le components

sub-

mission

assess-

ment

delivery

Proc.

phaseIndicator name Indicator values

availability


Validation


What kind of CRI distributions arise?

2014.12.04. 45

average

CRI

Per

winning

bidder

2009-

2012

Hungary


Political favoritism in Hungary

2014.12.04. 46

Hungary, total public procurement market, HU, 2009-2012

0%

10%

20%

30%

40%

50%

60%

70%

2009 2010 2011 2012

tota

l mar

ket

shar

es

"surprise" losers "surprise" winners

Election year


Political Control Indicator (PCI)

• Whether a company has direct political

connections

PCIi =

2014.12.04. 47

1, if company i has pol. connections

0, if company i does NOT have pol. conn.


Indicator validity

• The indicators co-vary

• CRI + PCI, HU, 2009-2012

2014.12.04. 48

Group N Mean CRI Std. Err. Std. Dev. 95% Conf.Interval

0=no political connection 2900 0.254 0.002 0.111 0.250 0.258

1=politically connected 1449 0.265 0.003 0.110 0.260 0.271

combined 4349 0.258 0.002 0.111 0.254 0.261

difference (CRI1-CRI0)

-0.011*** 0.004

-0.018 -0.004

0.011**


Indicator validity at EU level • Our indicators relate to external variables as expected: money

laundering, diversion of funds

• Financial Secrecy Index + CRI in EU, 2009-2013, N=26.309

2014.12.04. 49


Some illustrative findings


Applications - overview

Evaluating political change: Hungary before and after the general elections (2010), from socialist to Orban’s gov.

Evaluating countries: against each other or the same country over time

Evaluating large funding programmes: e.g. EU structural funds in CEE

2014.12.04. 51


Identifying focal points for policy intervention: e.g. networks of corruption

Evaluating the impact of single regulatory or organisational changes on corruption: e.g. loosening transparency regulations

Risk-based audit of actors/transactions

2014.12.04. 52

Applications - overview


CRI in Hungary before and after Orban’s government,

N= 43.642

0.1

0.2

0.3

0.4

0.5

0.6

20

09

01

20

09

02

20

09

03

20

09

04

20

09

05

20

09

06

20

09

07

20

09

08

20

09

09

20

09

10

20

09

11

20

09

12

20

10

01

20

10

02

20

10

03

20

10

04

20

10

05

20

10

06

20

10

07

20

10

08

20

10

09

20

10

10

20

10

11

20

10

12

20

11

01

20

11

02

20

11

03

20

11

04

20

11

05

20

11

06

20

11

07

20

11

08

20

11

09

20

11

10

20

11

11

20

11

12

20

12

01

20

12

02

20

12

03

20

12

04

20

12

05

20

12

06

20

12

07

20

12

08

20

12

09

20

12

10

20

12

11

20

12

12

upper-bound CRI (per contract)

gov't change

Application – Gov. Change


Application - EU Funding in CEE

• EU Funds increase corruption risks in Cz and Hu

(Ncz= 39.320, Nhu = 38.862)

• EU Funds decrease corruption risks in Sk (N=15.760)

2014.12.04. 54


Application - EU Funding in CEE

• Divergent temporal patterns

• Effect of new government in Hungary (2010)

2014.12.04. 55

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

20091 20092 20093 20094 20101 20102 20103 20104 20111 20112 20113 20114 20121 20122 20123 20124

me

an

CR

I

CZ:EU CZ:non-EU HU:EU HU:non-EU

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

20091 20092 20093 20094 20101 20102 20103 20104 20111 20112 20113 20114 20121 20122 20123 20124

me

an

CR

I

SK:EU SK:non-EU


Type of Issuers (hospitals): Share of PP with One Bidder by

Issuer Type, [Hungary, 1998-2004, N = 24.251]

Application – assessment at insititutional level motivation corrupion measurement data indicators illustrations plans

Applications - State capture

2014.12.04. 57

Captured

org.s’

network,

HU,

2009-

2010Q2

Socialist

gov.


Applications - State capture

2014.12.04. 58

Captured

org.s’

network,

HU,

2011-

2012Q2,

Orban’s

gov.


Further applications


From research to impact

• DIGIWHIST (H2020 project):

– The Digital Whistleblower. Fiscal Transparency, Risk Assessment and Impact of Good Governance Policies Assessed

– Consortium of Cambridge, Hertie, CRCB, Datlab, Open Knowledge Foundation, Transcrime

• Goals

– Help to anti-corruption efforts

– Open data and indicators for 35 European countries: EU, EEA, Caucasus

– Enabling losers of corruption to mobilize

2014.12.04. 60


• Data:

– Transparency and procurement legislation

– Micro-level procurement data

– Company information

– Public organisation information

– Asset declarations

• Indicators:

– Corruption

– Transparency

– Administrative quality

2014.12.04. 61

DIGIWHIST: key outputs


Further information Corruption Research Center Budapest: www.crcb.eu

Fazekas, M., Tóth, I. J. (2014), New ways to measure institutionalised grand corruption in public procurement. U4 Brief October 2014:9, U4 Anti-Corruption Resource Centre, Bergen, Norway.

Fazekas, M., Chvalkovská, J., Skuhrovec, J., Tóth, I. J., & King, L. P. (2014). Are EU funds a corruption risk? The impact of EU funds on grand corruption in Central and Eastern Europe. In Alina Mungiu-Pippidi (ed.) The Anticorruption Frontline. The ANTICORRP Project, vol. 2., (pp. 68–89). Berlin: Barbara Budrich Publishers.

Fazekas, M., Tóth, I. J. (2014), In respectable society: on how elite configuration influences patterns of state capture in Hungary. Conference paper, MPSA Annual Conference, Chicago, USA, 3 April 2014.

Fazekas, M., Tóth, I. J., & King, L. P. (2013). Anatomy of grand corruption: A composite corruption risk index based on objective data. CRC-WP/2013:02, Budapest: Corruption Research Centre.

Fazekas, M., Tóth, I. J., & King, L. P. (2013). Corruption manual for beginners: Inventory of elementary “corruption techniques” in public procurement using the case of Hungary. CRC-WP/2013:01,Corruption Research Centre, Budapest.

Fazekas, M., Tóth, I. J., & King, L. P. (2013). Hidden Depths. The Case of Hungary. In A. Mungiu-Pippidi (Ed.), Controlling Corruption in Europe vol. 1 (pp. 74–82). Berlin: Barbara Budrich Publishers.

2014.12.04. 62


http://www.crcb.eu/

Thank you for your attention!

2014.12.04. 63

Documents

Mihály Fazekas - István János Tóth