DiscreKon: definiKons
Cambridge English Dic2onary
the right or ability to decide something
Merriam Webster
the right to decide what should be done in a parKcular situaKon
Full defini2on
the quality of having or showing discernment or good judgement ability to make responsible decision
Good, if not abused
Necessary in important situaKons
Of course, discreKon may be abused
Need for accountability
Cambridge English Dic2onary
a situaKon in which someone is responsible for things that happen and can give a saKsfactory reason for them
No discreKon, no decision, no responsibility, no accountability
Rules limit discreKon, hence may limit its abuse
but also limit decisions, responsibility, and accountability
Pros of Rules
• By limiKng discreKon/decisions, they reduce need for ex post accountability controls
• Cheaper ex ante “procedural/bureaucraKc” controls instead
• Simplify tasks, reducing needs for judgment, training, and cost of personnel
Cons
• Lead to poorer outcomes: rules are “stupid”, rigid, do not adapt to local info/specific condiKons
• DemoKvate skilled civil servants: just follow rules! Poor selecKon, moKvaKon, quality…
• May confuse responsibility/limit accountability
In a very corrupt state, there will be lots of laws (Tacitus)
CompeKKon as a Rule: Use Open AucKons
1. Reduce discreKon in choosing contractors/risk of its abuse 2. Introduces new watchdogs: compeKtors
Plus
Promise of large gains in Value for Money (max Q/P), at least as measured at the awarding stage
True if contracts almost complete, well wriben, well coordinated with the scoring rule, well managed…
Limits to CompeKKon/Open AucKons
• Complete contracKng condiKons are very ocen not met, as the private sector shows (Toyota vs GM… Kelman 1990)
• Apparent gains in ex ante offered Q/P may correspond ex post to very low Q, or renegoKaKon, or corrupKon
• Monitoring Q much more difficult than P, hence all under the lamp looking at P alone… guess the results?
Buyer DiscreKon in Dynamic Procurement
If performance hard to contract, “quality depends on price” and standard arguments for compeKKon fail (SKglitz 1989, Spulber 1990, Manelli e Vincent 1995)
• In repeated procurement relaKonships, limited contractual governance ‘completed’ by relaKonal-‐reputaKonal forces – RelaKonal: Vendor RaKng systems
– ReputaKonal: Brand, Feedback mech. (eBay)
• Rela2onal governance needs discre2on, acKng on non verifiable informaKon, also to fight corrupKon! (Calzolari-‐Spagnolo 2009)
CompeKKon as a Rule: Use Open AucKons
When important quality aspects not verifiable, only low quality ends up being purchased, and fall in Q may dominate that in P AlternaKves:
• NegoKaKons:
introduce some discreKon but almost eliminate compeKKon allows higher quality but at increasing cost
accountability checks very difficult/expensive, easy to abuse
• Restricted aucKons by invitaKon:
introduces discreKon (ability not to invite some suppliers) maintains substanKal compeKKon, less easy to abuse
PosiKve Evidence on DiscreKon in Public Management
Rasul and Rogger (2016), study Nigerian Civil Service, find that
• increasing bureaucrats’ autonomy is posiKvely associated with compleKon rates,
Duflo et al. (2015) field experiment on environmental regulaKon and its enforcement in India: finding, among other things, that
• regulatory discre2on is highly valuable, it allows the regulator to beber target inspecKons at extreme polluters,
Bandiera et al. (2009): waste in the procurement of Italian goods and services is substanKally smaller for more autonomous public purchasing authoriKes that enjoy more discreKon.
Evidence on AucKons vs NegoKaKons in Public Proc.
Vellez (2011) on Italian Health procurement, mulKlateral
compeKKve negoKaKons, no difference in price with aucKons
Chever and Moore (2012) on construcKon of housing in France.
They find that negoKaKons acer an informal aucKon are associated with lower costs relaKve to open aucKons.
Lalive, Schmutzler and Zuelenher (2015): procurement of rail services in Germany, hand collected data
• aucKons improve surplus (service & prices) of 30%;
• surplus would have increased by 16% on negoKated lines;
• administraKve costs of carrying out aucKons.
Evidence on Open vs Restricted AucKons in Public Proc.
Beber comparison to isolate the effects of discreKon, as more compeKKon is maintained.
Pros
Coviello, Guglielmo e Spagnolo (2011/2016) analyze works in Italy 2000-‐2006 using EU threshold, causal evidence (RDD):
• DiscreKon (> restricted but compeKKve aucKons, min 15 invited) increases repeated buyer-‐seller interacKon
• Does not worsen outcome (prices, delays, renegoKaKons)
• For some specificaKons, it seems to improve delivery Kme AddiKonal evidence (propensity score matching)
• well performing, large, non-‐local firms selected again
Good effects of discreKon appear to at least compensate bad
Open vs Restricted AucKons in Public Proc. cont’d
More ProsChever, Saussier and Yvrand-‐Billon (2014) study 180 contracts by a local public buyer of social housing 2006-‐2009, show that
• restricted aucKons lead to same outcomes as open but lower transacKon costs, and no signs of abuse of discreKon
Kang and Miller (2015) esKmate proc. model where extent of compeKKon is opKmally chosen by public buyers, and find that
• limiKng compeKKon need not result in higher proc. costs.
Con
Gurakar and Meyerson (2016): increased use of restricted aucKons in Turkey works led to
• higher costs and
• higher winning probability for AKP-‐affiliated firms
Open vs Restricted AucKons in Public Proc.: Cons
Palguta and Pertold (2013), Check proc. data, EU reform 2006
• Restr. Auct. è value manipulaKons to stay below threshold
• Outcomes: anonymous firms win more ocen, at higher prices Hypothesis
Illegal discreKon (through strategic bunching) has worse effects?
Transparency and data for ex post performance evaluaKon are essenKal to foster the posiKve effects of discreKon (see Saussier & Tirole 2015)
example
Public Procurement Thresholds and Data in Sweden
Report for Konkurrensverket by SITE-‐Stockholm School of Economics
Team: Roman Bobilev, Andrea Guglielmo, Elena Paltseva, Giancarlo Spagnolo
– An exploratory study of Public Procurement in (Italy and) Sweden which focuses on:
(i) strategic ”bunching” of tenders below regulatory thresholds, with illegal increase in discreKon;
(ii) the heterogeneity of contracKng authoriKes’
organizaKon and behavior in public procurement;
(iii) procurement data availability, also in comparison to some other European countries.
Ch. 2
Bunching and Outcomes: and example from Italian procurement of public works
Discre0on and Public Procurement Giancarlo Spagnolo
• Database from the former Italian Authority for the Surveillance of Public Procurement (AVCP)
• All the public works awarded in Italy between 2000 and 2005 with values greater than or equal to 150,000 Euro
• For each contract, we observe the number of bidders, the winner's rebate, the project value, the idenKty of the
winning bidder, the type of work, the final cost, the date of delivery, and the type and locaKon of the public
administraKon
• Focus on a subsample between 200k and 500k Euro.
Data as in Coviello, Guglielmo e Spagnolo (2011)
Approach
• Focus on heterogeneity in sorKng behavior across different contracKng authoriKes, and the impact of it on
procurement outcomes
• First, we compare the behavior of a “Big Player” to other contracKng authoriKes
• Second, we compare the contracKng authoriKes according to their level of corrupKon, social capital and judicial
efficiency
Aggregate data – not much apparently
Taking heterogeneity into account…
StaKsKcs for “Big Player”
“Big Player” vs. other administraKons:
• Fewer delays
• Fewer deviaKons from the iniKally planned costs
• Smaller probability to award to local bidders
• Higher probability of previous suppliers winning bids
• “TrabaKva Privata” used more below threshold (20% vs 9%)
• Less delays and same costs with “TrabaKva Privata”
Other staKsKcs
• Areas with higher corrupKon
– More bidders and more substanKal rebates
– “TrabaKva Privata” is used less ocen below threshold
• Areas with higher social capital
– More bidders and more substanKal rebates, no maber with what procedure or value
– “TrabaKva Privata” is used more ocen below threshold
• Areas with beber judicial efficiency – Fewer rebates and fewer bidders
– Past winners win more ocen, fewer delays
Conclusion
• “Big Player” likely to bunch and use discreKon more ocen
• “Big Player” more efficient, beber results
Consistent with posi2ve use of illegally obtained discreKon
• Use of open aucKons and (apparent) compeKKon not associated to beber performance and less corrupKon (post-‐aucKon phase appears therefore crucial…)
• The type of administraKon plays a crucial role
Ch.3
Accessing bunching below threshold in Sweden
Discre0on and Public Procurement Giancarlo Spagnolo
Sweden
Can we assess the effect of the thresholds in Sweden?
• Specifically, we look at the EU Public Procurement direcKves -‐
Procurement tenders with the expected value above the threshold should be procured according to EU-‐wide rules
– Contract noKces published at EU-‐wide website TED (Tenders Electronic Daily)
– More detailed data requirements, etc.
• Main ques0on: Is there an evidence of strategic behavior in procurement due to thresholds? Is there impact on
procurement outcomes? Does it depend on the type of the buyer?
Public procurement data in Sweden
Swedish procurement data on tender value:
• Above EU thresholds – TED database
– Compliance is very poor, see next slide – Expected tender value cannot be uKlizes
• Below EU thresholds -‐ VISMA database
– Covers around 80% of public procurement deals in Sweden – Also includes the data reported through TED
– However, expected value is NOT collected
• it may sKll be somewhere in pdf documents, but not for systemaKc analysis
– Also, data is on tenders, not on individual lots – difficult to measure compeKKon
– Outcome data is not collected
• Underreported for both es0mated and actual value, also in comparison to other EU member states.
• IdenKty of buyer mabers: Swedish public enKKes more likely to report the values are those
• in the central administraKon and public uKliKes sectors,
• procuring for the goods and services within the financial and insurance services and architecture and construcKon services industries.
• The data on es0mated value also looks non-‐reliable
– likely due to misunderstanding or different interpretaKon of the EU regulaKon
• Example: distribuKon of winner rebate
• BoBom line:
– EsKmated value data for Sweden cannot be used for empirical analysis
– There is a need in clarificaKon of the EU procurement regulaKon to public enKKes in Sweden
– Perhaps, also a need for beber alignment between the EU legislaKon and different cross-‐country procurement pracKces .
• Data issues (e.g., we have to use final, as opposed to esKmated, procurement value)
– Our measures of bunching are imprecise
• SKll, we find signs of bunching under the EU thresholds!
– Confirmed by both histograms and McCrary density disconKnuity tests (see next slides)
• Bunching depends again on the idenKty of the buyers
– Supplies and services: staKsKcally significant bunching for central governments
– Public works: (close to significant) bunching for other authoriKes
• data here incomplete: many of public works procurement tenders – of Traffikverket -‐ are not in VISMA
Bunching in Sweden
• Histogram analysis
• McCrary density disconKnuity tests
• BoBom line:
– Evidence of strategic bunching in public procurement in Sweden
– More and beber data would be needed for a thorough analysis of the effect of this bunching
Conclusions
• Increase in discreKon may have good or bad effects, depending on insKtuKonal environment (France-‐Italy vs Turkey)
• Rules constraining discreKon easily avoided, then illegal gain in discreKon, for the good or for the bad (Italy vs Check Republic)
• Same happening in Sweden, but no data to evaluate effects discreKon needs data for ex post accountability checks!!
• Different types of contracKng authoriKes behave rather differently in public procurement.
• Complex problem, simple rules subopKmal in most cases.
• Transparency and data for ex post performance evaluaKon are essenKal to foster the posiKve effects of discreKon (Saussier &
Tirole 2015)
