Alessio D'Ignazio

Spatial Dispersion of Peering Clusters in the European Internet

We study the role played by geographical distance in the peering decisions between Internet Service Providers. Firstly, we assess whether or not the Internet industry shows clustering in peering; we then concentrate on the dynamics of the agglomeration process by studying the effects of bilateral distance in changing the morphology of existing peering patterns. Our results show a dominance of random spatial patterns in peering agreements. The sign of the effect of distance on the peering decision, driving the agglomeration/dispersion process, depends, however, on the initial level of clustering. We show that clustered patterns will disperse in the long run.

'Unfair' Discrimination in Two-sided Peering? Evidence from LINX

Does asymmetry between Internet Providers affect the “fairness” of their interconnection contracts? While recent game theoretic literature provides contrasting answers to this question, there is a lack of empirical research. We introduce a novel dataset on micro-interconnection policies and provide an econometric analysis of the determinants of peering decisions amongst the Internet Service Providers interconnecting at the London Internet Exchange Point (LINX). Our key result shows that two different metrics, introduced to capture asymmetry, exert opposite effects. Asymmetry in “market size” enhances the quality of the link, while asymmetry in “network centrality” induces quality degradation, hence “unfairer” interconnection conditions.

Spatial Dispersion of Interconnection Clusters in the European Internet

Abstract This paper studies the effects of geographical distance on the interconnection agreements between providers participating at Internet Exchange Points (IXPs) in Europe. We assess separately two main issues: the extent of existing clustering, as well as the role of proximity in bilateral interconnection decisions. Our results show the dominance of spatial random patterns in the interconnection agreement distributions. On the other hand, we find that proximity positively affects the probability of establishing interconnection for all the IXPs studied but one. Interesting, the latter is the only one showing an initially clustered morphology. This indicates a prevalence of centripetal forces—local spillovers and mutual knowledge—over geographical differentiation for any pre-existing interconnection pattern morphology except the clustered ones.

Antitrust Analysis for the Internet Upstream Market: A BGP Approach

In this paper we study concentration in the European Internet upstream access market. Measurement of market concentration depends on correctly defining the market, but this is not always possible as Antitrust authorities often lack reliable pricing and traffic data. We present an alternative approach based on the inference of the Internet Operators interconnection policies using micro-data sourced from their Border Gateway Protocol tables. Firstly we propose a price-independent algorithm for defining both the vertical and geographical relevant market boundaries, then we calculate market concentration indexes using two novel metrics. These assess, for each undertaking, both its role in terms of essential network facility and of wholesale market dominance. The results, applied to four leading Internet Exchange Points in London, Amsterdam, Frankfurt and Milan, show that some vertical segments of these markets are extremely competitive, while others are highly concentrated, putting them within the special attention category of the Merger Guidelines.

Public guarantees on loans to SMEs: an RDD evaluation

The paper evaluates the impact of the guarantees provided by the Italian scheme Fondo di Garanzia on the access to credit for small and medium enterprises. The study exploits the mechanism that assigns the guarantees, which is based on a scoring system to assess eligibility. By using regression discontinuity techniques, the paper finds that at the threshold between eligible and non-eligible firms, the program has a positive impact on bank loans to firms; however, the scheme has no impact on the interest rate charged by the banks, while it affects positively the likelihood that a firm becomes unable to repay its loans. The guaranteed loans were mostly used to finance working capital. Finally, the paper provides inference for far-from-the-threshold firms.

Targeting policy-compliers with machine learning: an application to a tax rebate programme in Italy

Machine Learning (ML) can be a powerful tool to inform policy decisions. Those who are treated under a programme might have different propensities to put into practice the behaviour that the policymaker wants to incentivize. ML algorithms can be used to predict the policy-compliers; that is, those who are most likely to behave in the way desired by the policymaker. When the design of the programme is tailored to target the policy-compliers, the overall effectiveness of the policy is increased. This paper proposes an application of ML targeting that uses the massive tax rebate scheme introduced in Italy in 2014.

Clustering, Connectivity and Power Laws: Useful Tools for Assessing Market Structure of the European Internet?

We study the relation between connectivity and clustering, of Internet Service Providers, ISPs, to understand the hierarchy within European Internet Exchange Points, IXPs. We assess the possibility of power law relations between clustering and connectivity, typical markers for a hierarchical Internet market structure. We find a predominantly inverse relation between local clustering and peering strength confirming that, within the members of a given IXP, the most connected ISPs have direct neighbours that are poorly connected among themselves. The implications for market analysis of these studies of the Internet Connectivity at are discussed.