We investigate the impact of procurement thresholds on strategic behavior of public buyers in Sweden. We document signs of “bunching” at the threshold, which suggests that strategic behavior in procurement is potentially important in Sweden, and should not be overlooked in the on-going public debate on the procurement thresholds. At the same time, data limitations do not allow us to access the impact of this strategic behavior on procurement outcomes and efficiency. This calls for better and more extensive procurement data collection.
Thresholds in Public Procurement Regulation
The impact of public procurement regulation on the incentives of public buyers and resulting procurement outcomes has recently become the focus of both political and academic debates. This interest is not surprising. Public procurement has been playing an important role in economic life. For example, the annual value of public procurement purchases in the EU is around 2.5 trillion Euros, which constitutes about one fifth of the EU GDP (with figures for individual Member States ranging from 9 to 30%, EC (2012)).
Many national and international procurement regulations specify different procedures and requirements for deals with an estimated value exceeding/falling short of certain “thresholds”. For example, the EU Public Procurement Directives establish that national tenders with expected value above certain thresholds should be publicized at EU level and procured in line with the common EU-wide rules. Regulations for above-the-threshold procurements are typically associated with stronger demands on the transparency of the procurement procedure and a more rigorously regulated procurement format. As a result, there is a potential incentive for a public procuring entity to manipulate the value of the tender to stay below the threshold and avoid these additional demands.
This manipulation does not imply negative consequences for the procurement outcomes per se. Public buyers may use the additional discretion arising from setting the value below the threshold to avoid administrative costs, and to make the procurement process faster and potentially more efficient. At the same time, public buyers may use the looser regulation below the threshold to reduce supplier competition (e.g., by limiting tender advertising), or to facilitate corrupt procurement deals with specific pre-chosen contractors. In either case, the thresholds can potentially give rise to strategic behavior of the public procurers.
However, little is known empirically about the extent of this behavior, and, more generally, about the impact of procurement thresholds on procurement outcomes. Coviello and Mariniello (2014), and Coviello, Guglielmo and Spagnolo (2014) use regulatory thresholds to study the effects of regulation – advertising rules and ability to use restricted auctions, respectively – on the outcomes of public procurement in Italy. The latter study also finds evidence of value manipulation around the threshold but only in some industries, such as road construction. Palguta and Pertold (2014), and Jascisens (2014) document manipulation of project value to avoid thresholds for Czech Republic and Latvia respectively, and show that this manipulation is likely to be associated with corrupt behavior and favoritism.
In the case of Sweden, public procurement regulation sets two types of thresholds. First, it stipulates that all public purchases above a certain relatively low threshold must be announced in a local public database, and prescribes the procurement procedures for these purchases. In turn, it requires that the purchases exceeding the EU-wide (and higher) thresholds should be managed according to the above-mentioned EU Directive rules.
There is not much known empirically on the effect of thresholds in Swedish procurement. The recent Swedish debate on where to set procurement thresholds has mostly addressed the issue of the administration costs associated with more transparent procedures above the threshold(s), see e.g. Molander (2009, 2013). However, the above arguments and evidence calls for an investigation of strategic consequences of the thresholds in Sweden.
Thresholds and Strategic in Procurement Behavior in Sweden
In this brief, we summarize the results of our study (Bobylev et al. (2015)) aimed at assessing threshold-driven strategic behavior of public buyers in Sweden.
We use OPIC, the proprietary database covering most of the Swedish public procurement tenders, kindly provided by its owner and maintainer, Visma Commerces AB (hereafter Visma), to see whether there are any signs of ”bunching” – i.e. of unusually high concentration – of Swedish tenders below the EU-wide procurement thresholds. The database contains information both on tenders above and below the EU-threshold; however, there are certain difficulties in using these data for the above-mentioned purpose. In particular, the EU Public Procurement Directives thresholds, and the corresponding part of the Swedish procurement regulation, are explicitly formulated in terms of the estimated value of the procurement tenders. Thus, value manipulation behavior of public buyers should be revealed by their choice of the expected procurement value around the threshold. However, OPIC (or any other database on Swedish procurement) does not contain/collect this estimated value for neither below- nor above-threshold deals. We circumvent this complication by basing our analysis on the actual, realized tender value, which should be highly correlated with the estimated value.
We start the analysis by providing illustrative histograms of tender values around the relevant thresholds. Figures 1 and 2 illustrate the distribution of tender value for procurements in supplies and services, and in public works, respectively. The sample is also split by the type of public buyer: central authorities vs. other public entities.
Figure 1. Histograms of Tender Values, Supplies and Services
Figure 2. Histograms of Tender Values, and Public Works
The red vertical line corresponds to the EU procurement threshold (which differs for procurement in supplies and services, and in public works, and across the central vs. non-central authorities).
These histograms suggest that there are signs of bunching below threshold – for example, the graph for central government procurement in supplies and services shows a spike just below the threshold, and that for public works and other contracting authorities seems to present irregular drop just after the threshold.
We proceed by formally testing for the presence of distribution discontinuity at the threshold by using a density test technique as in McCrary (2008). Effectively, this test smoothens the histogram using linear regression separately to the left and to the right of the discontinuity point. If we detect a discontinuity, this may be indicative of systematic manipulation of procurement values by public buyers.
Figure 3. McCrary Density Tests for Supplies and Services, and Public Works
The graphs in Figure 3 present our resulting smoothers and associated 90% confidence intervals (which corresponds to testing the bunching hypothesis at 10% significance level). We normalize the value of tenders attributed to each of the thresholds at the threshold itself; therefore, discontinuity, if any, should be expected at zero.
The results of the test are in line with earlier histograms. There is certain discontinuity in the procurement distributions; it is especially evident, and statistically significant at 10%, in procurement of supplies and services undertaken by central government buyers. In other words, we find significant strategic bunching, although non-universally across the types of public buyers/types of procurement.
The next logical step would be to analyze the effects of this strategic bunching at the threshold on the outcomes of the public procurement, such as the quality of the tenders or competitiveness of the process. However, systematic data on procurement outcomes in Sweden are currently not available, preventing us from accessing the consequences of this bunching.
We investigated the effect of thresholds on public procurement in Sweden, and have found signs of strategic bunching of public buyers at the EU thresholds. In other words, procurement thresholds affect not only the associated transaction costs, but also the strategic behavior of procuring entities, a point overlooked in the current Swedish debate on public procurement.
This strategic behavior may potentially have serious implications for the procurement outcomes, and, more generally, societal well-being. However, current limitations of data made us stop short of investigating these implications. The above arguments thus call for more and better procurement data collection in Sweden.
- Bobilev, R., Guglielmo, A., Paltseva, E. and Spagnolo, G. (2015). ”Public Procurement Thresholds and Data in Sweden”, KKV Report 2015:3, http://www.kkv.se/globalassets/publikationer/uppdragsforskning/forsk_rap_2015_3.pdf
- Coviello, D., and Mariniello, M. (2014). “Publicity requirements in public procurement: Evidence from a regression discontinuity design.” Journal of Public Economics 109, pp. 76-100.
- Coviello, D., Guglielmo, A., and Spagnolo, G. (2014). “The Effect of Discretion on Procurement Performance“ Mimeo
- European Commission (2012), “Public Procurement Indicators 2011”,
- Jascisens, V. (2014). “To Divide or Not – Threshold Effects in the Below Threshold Procurement in Latvia“, Master Thesis, 2012, Toulouse School of Economics
- Molander, P. (2009). “Regelverk och praxis i offentlig upphandling”. Rapport till Expertgruppen för studier i offentlig ekonomi, 2009:2. Finansdepartementet.
- Molander, P. (2013). ”Höjd nivå för upphandling blir dyrt för skattebetalarna”. Dagens Nyheter 11 May
- Palguta, Ján, and Filip Pertold (2014). “Corruption and Manipulation of Public Procurement: Evidence from the Introduction of Discretionary Thresholds.” Mimeo