Shoot the Messenger: Statistical Credibility, Institutional Independence, and Sovereign Risk Premia

Governments that dislike their own statistics increasingly go after the people and institutions that produce them. This paper asks what such attacks cost in the sovereign bond market, using the most extensively litigated case on record: the decade-long criminal prosecution of Andreas Georgiou, the head of Greece's statistical authority, for certifying the country's true 2009 deficit. I combine daily Eurostat data on ten-year sovereign yields with a chronology of seven verifiable legal milestones between 2011 and 2018, and contrast them with the 2009–2010 episodes in which the misreporting itself was revealed. Three findings emerge. First, markets priced the lie: the January 2010 European Commission report on Greek statistical irregularities alone is associated with an abnormal spread increase of 0.74 percentage points within ten trading days, and the Greek-specific repricing during the revelation phase reached 3.5 percentage points, a present-value cost in the tens of billions of euros. Second, markets did not price the trial: across every prosecution escalation and acquittal, daily abnormal spread movements are economically small and statistically indistinguishable from zero, a null that survives a two-way fixed-effects panel design, donor-pool permutations, alternative event dates, equity-market proxies, and an extensive specification battery. Third, the two results are connected: because Eurostat certified the corrected statistics repeatedly from November 2010 onward, the prosecution carried no information about data quality, and an externally anchored credibility is precisely what the market had already priced. The policy implication is uncomfortable for those who expect market discipline to police institutional quality: bond markets penalize governments that manipulate data, but they provide no measurable deterrent against the persecution of the statisticians who correct them. Protection of statistical independence must therefore come from institutional design, not from creditors.