Calendar
Newsletter
The Portuguese Public Finance Council (CFP) publishes this report to assess the quality of the State Budget forecasts prepared by the Ministry of Finance (MoF). Therefore, the CFP implements the provisions of Article 8(3) and 8(4) of the Budgetary Framework Law (Law 151/2015 of 11 September), as amended, which transposed Article 4(6) of Council Directive 2011/85/EU into national legislation.
This study carries out a quantitative and statistical analysis of the Ministry of Finance's macro-budgetary forecast deviations, by calculating different performance measures and trying to assess whether these forecasts fulfil some of the optimal properties - unbiased, no autocorrelation, accuracy and weak-efficiency.
The analysis covers the period 2015 to 2022, sample for which there are macroeconomic and fiscal forecasts made under ESA 2010. Excluding the pandemic year (2020), the real-time results show that the forecast's optimal properties are met for most of the analysed variables, with a few exceptions in the case of unbiasedness and weak-efficiency:
(i) with regard to unbiasedness, there is statistical evidence that only the forecasts for employment growth and the ratios to GDP of total expenditure and interest expenditure are biased;
(ii) with regard to the absence of autocorrelation, this hypothesis is only rejected in the case of the employment forecasts (autocorrelation of order two); and
(iii) with regard to weak-efficiency, the tests reject this hypothesis in the case of the forecasts for employment growth, and the expenditure, revenue and interest ratios. All these tests are compared against the conventional significance levels (1%, 5% and 10%).
As far as accuracy is concerned, the Diebold-Mariano test shows that the MF forecasts are not statistically more accurate than the naïve forecasts, i.e. considering the last observed value in real-time. However, given the small size of the sample, these results should be interpreted with particular caution. It should also be noted that carrying out this analysis with final data instead does not significantly change these conclusions, except in the case of the forecasts for real and nominal GDP, which are now considered biased. This difference results from the magnitude of the revisions of the annual National Accounts between the first estimate (real-time) and the final data.
In short, the tests carried out lead us to conclude that the MoF's forecasts can be considered prudent, although a larger sample is needed to obtain more robust conclusions.
Date of last update: 31/01/2024
