In a recent VoxEU column, Beetsma et al (2011) highlight “Europe’s fiscal policy credibility problem”. They show how fiscal plans get steadily worse as governments move from planning to implementation, even in the case of estimated outcomes published towards the end of the year being forecast. They stress how this behaviour harms the credibility of fiscal policies in Europe and how appropriate fiscal institutions could play a substantial role in correcting these problems, especially if EU countries were to take more ownership of EU fiscal rules.
Do governments report credible actual (available in real-time) figures?
Against this background, we analyse a related issue that turns out to be crucial for the credibility of fiscal policy plans. Do governments report credible actual (available in real-time) figures? In the EU, actual figures for a given year t are reported in the spring of year t + 1 for the first time and, according to the EU’s statistical legislation, can be revised subsequently up to eight times in the bi-annual reports. During this process of revisions, a year that initially presented a fiscal surplus may end up showing a deficit. But this may occur after a number of years, possibly when a different government is in office and with such a lag that the fiscal situation of the year being revised is of no interest to the markets or the public at large.
A recent example is the Greek case. The international press has been echoing, especially since October 2009, the strong and recurrent revisions to the Greek fiscal figures. Indeed, even though Greek fiscal figures have been under scrutiny by the European Commission (Eurostat) since at least the early 2000s, renewed attention is being paid to this case. Figure 1 illustrates the successive revisions to Greek fiscal deficit figures reported over the period 1999-2009, and shows how revisions in key years, like the ones being used to judge accession to the Eurozone, have been wildly revised ex post.
Figure 1. Greek public deficits (as a share of GDP) as published in real-time
Recently, some voices have been asking whether revisions to Greek fiscal data are representative of Europe’s fiscal reporting system or, conversely, are just an isolated occurrence that can be explained by Greece’s idiosyncratic factors. According to the Financial Times (2010) “the Commission said EU fiscal data were generally of high quality and Greece represented a one-off problem. However, it cautioned that it lacked audit powers and so relied heavily on the goodwill and integrity of member-states to supply accurate data”. In the same fashion, the President of the ECB recently stated that “while the government finance statistics of the overwhelming majority of the Member States is reliable, this does not yet apply to all of them”.
The reporting of fiscal data is at the heart of the EU’s multilateral surveillance rules. Adherence to such rules is judged upon initial data releases by EU Member States, in the framework of the EDP Notifications, one of the bases of Europe’s Stability and Growth Pact. Under the Pact, any deviation of government deficits from the reference value of -3% of GDP in the base year (not considered to be corrected over a given forecast horizon) leads to the adoption of corrective actions. In this case, the country concerned has to engage in a path of deficit reduction to be approved by the Council of Ministers of the EU. Frequent and/or sizeable revisions of fiscal data may give rise to concerns about the reliability of the official statistics used in the monitoring.
Even though the production of fiscal data in Europe is the responsibility of independent national statistical institutes and is subject to scrutiny by the European Commission - more precisely by Eurostat, the EU’s statistical agency - from a political-economy point of view, some governments might have incentives to resort to creative accounting practices so that initially released figures are distorted (see Kohen and Van de Noord 2005). In addition, when sizeable fiscal data revisions become a stylised fact, the credibility of government targets is also at stake. Where revisions do not present a clear recurrent pattern but are nonetheless frequent, the comparison of successive paths of government targets might be blurred, and may eventually undermine the soundness and consistency of fiscal policy choices over time.
Patterns of past fiscal data revisions in Europe
In a new paper (De Castro et al 2011), we look at publicly available Eurostat press releases published between spring 1999 and October 2009 for 15 EU countries - Belgium, Germany, Greece, Spain, France, Italy, Luxembourg, Netherlands, Ireland, Austria, Portugal, Finland, Denmark, Sweden and the UK. In Table 1, we show the mean of public deficit revisions at different points in time. Within the first year, there are no revisions on average. However, as time goes by, a negative revision bias emerges that becomes significant after three years, and reaches a maximum after four years (the final figure, according to EU rules). Two comments are warranted, (i) revisions show successively higher public deficits, and (ii) this bias is only significant after three years. Clearly, from today’s perspective the publication of a revised 2008 public deficit figure is not going to affect the public debate or the voting decision of deficit-averse citizens.
Table 1. Fiscal data revisions
|Revisions within the 1st year||Revisions after 1 year||Revisions after 3 years||Revisions after 4 years|
|No. of observations
Note: * The mean for the Pool is significant at the 1% level.
The aggregated figures of Table 1 hide some cross-country heterogeneity (with Greece being clearly an outlier) and also some heterogeneity along the business cycle. The data reveal that the negative bias is highest between 2001 and 2003, a period characterised by economic slowdown. By contrast, “good times” (1998-1999, 2005-2007) tend to present either positive or non-significant total (or partial) revisions.
Do fiscal data revisions add news or noise?
In this respect, a first step would be to discern to what extent revisions reflect the relevant “news” embedded in the newly available information, or rather reflect “noise” in the production of the previous estimate. In the case of “noise”, it is assumed that the revision does not contain any new information, while in the case of “news”, the change in the estimate of the variable of interest from one vintage to the next can be attributed to the incorporation of new information. The usual tests in this literature follow the seminal paper of Mankiw and Shapiro (1986). An application of these standard tests shows that the fiscal deficit revision process is a mixture of “noise” and “news”, with the statistical agencies only partially using the available information at each point in time. In general, these results suggest that the production process of fiscal data in the sample under study is not efficient.
Political-economy determinants of past fiscal data revisions
Beyond the information contained in previous vintages and the presence of systematic biases, other explanatory factors may help in understanding the reasons behind government deficit revisions, in particular the economic situation as measured by real GDP growth forecasts, and the proximity of elections. Our estimates confirm that these variables do matter in explaining deficit revisions. Specifically, if the cyclical situation is expected to be adverse and/or if a general election is approaching, governments appear to have incentives to offer an overly optimistic view of the state of public finances that leads to upward revisions of public deficits in the future. More concretely, our results show that governments have incentives to hide part of the actual deficit in a pre-election year, and to revise it afterwards, once the election has taken place. By contrast, newly appointed governments, especially after changes in its sign, tend to be overly pessimistic regarding deficit figures for the previous year, maybe because they can blame the previous administration. Accordingly, insofar as these variables have forecasting power on future revisions, the result of lack of rationality of initial data releases is reaffirmed.
Can fiscal rules be of any help? Some policy conclusions
To assess their role in the revision process, we include the European Commission’s fiscal rules index in the analysis. Although fiscal rules by themselves are unlikely to fully eliminate the bias towards an increased deficit witnessed in data revisions, the empirical results obtained show that more stringent fiscal rules could contribute to the reduction of the overall bias. Accordingly, our results support the view that stronger and more binding fiscal rules also contribute to enhancing the credibility of (earlier) fiscal data releases. According to EU statistical rules, the elaboration and provision of past fiscal data remains a responsibility of each individual country. An immediate implication of the outlined work is that EU countries should take more ownership of EU fiscal rules in this regard.
However, there is also some work to do on the statistical front. In the EU framework, Eurostat is mandated to observe whether the data reported by individual EU member states complies with the extant statistical guidelines. As a result of the scrutiny, a given country can be asked to revise their fiscal data. Now, one may ask whether Eurostat has asked countries to revise initially reported figures in one particular direction (for example towards more public deficit), or whether the outcome has been more balanced? To answer this particular question, we compiled all the revisions stemming from the implementation of Eurostat’s decisions on preliminary country data and on Eurostat’s clarifications on the interpretation of the national accounts’ rules. Interestingly, the vast majority of Eurostat-induced public deficit revisions led to higher deficits on average in the sample under study. This fact would be consistent with two opposite views. On the one hand, one may claim that a large share of the revisions were due to the still-evolving nature of the European statistical system; alternatively, it could also be claimed that individual countries tried to manipulate information in preliminary publications of data, by stretching to the limit the common accounting rules. Accordingly, closer monitoring of accounting practices by Eurostat, jointly with further efforts to clarify borderline issues, are warranted to strengthen the overall credibility of the multilateral fiscal surveillance mechanism.
Beetsma, R., B. Bluhm, M. Giuliodori, and P. Wierts (2011), “Europe’s fiscal policy credibility problem”, VoxEU, 1 July.
Beetsma, R., B. Bluhm, M. Giuliodori, and P. Wierts (2011), “From First-Release to Ex-Post Fiscal Data: Exploring the Sources of Revision Errors in the EU”, CEPR Discussion Paper 8413.
Financial Times (2010), Greece condemned for falsifying data, 12 January.
Kohen, V. and P. van de Noord (2005), “Fiscal Gimmickry in Europe: One-Off Measures and Creative Accounting”. OECD Economics Department Working Paper 417.
De Castro, F., J.J. Pérez and Marta Rodríguez-Vives (2011), “Fiscal data revisions in Europe”. European Central Bank Working Paper 1342.
Mankiw, N.G. and M.D. Shapiro (1986), “News or Noise? An analysis of GNP revisions”. NBER Workin Paper 1939.