Sovereign Ratings

For ease of reading, the rating notation from Fitch/S&P is used in the paper for all four rating agencies (e.g. we use ‘AA+’ rather than ‘AA(H)’ from DBRS or ‘Aa1’ from Moody’s).

The multiple notch decisions were the following: from Moody’s, −3 notches on 4 October 2011 and −2 notches on 13 July 2012; from Fitch, −2 notches on 27 January 2012; from S&P, −2 notches on 13 January 2012.

The Eurosystem applies the ‘first best rating rule’, therefore the change of CQS occurs when the first best rating (or the only available rating) moves to a different CQS.

We performed our analysis on the methodologies in force in 2020. Up to January 2023, credit rating agencies have performed only technical updates which do not affect the substance and conclusions of our work. More precisely, the reference documentation for our analysis is the following: DBRS Morningstar (2020a, 2021), Fitch (2020l, 2021), Moody’s (2019, 2021), S&P (2017, 2021b).

Moody’s, in particular, explains that expressing rating in cardinal terms would imply a higher rating volatility, while S&P explains that only in the long term would default frequencies be similar across similarly rated issuers from different sectors. See Moody’s (2021), p 37. See S&P (2021b) p 57.

See DBRS Morningstar (2020a) p 3, p 19; Fitch (2020l) p 7; S&P (2017) pp. 4–5, p 33.

Regulation (EU) No. 462/2013 of the European Parliament and of the Council of 21 May 2013 amending Regulation (EC) No. 1060/2009 on credit rating agencies.

The independent variables of the regression equation are the 18 economic and financial indicators of the model; the estimated coefficients of the regression are the same for all sovereigns and represent the weights of the variables in the equation. The dependent variable is a numerical score that provides, via a proprietary matching table, the model rating for the sovereign under assessment (e.g. a value between 6.5 and 7.5 corresponds to BBB-). The regression is estimated from the application of Ordinary Least Squares (OLS) to the set of the 18 variables for all sovereigns rated by Fitch over 2000–18 inclusive and is re-estimated and reviewed annually to incorporate additional data into the estimation period and to test for new potential variables. See Fitch (2020l), p 6.

See DBRS Morningstar (2020a), p 27.

See Fitch (2020l), p 23, footnote 1.

The adjustments, within +/−2 notches, are applied to the score assigned by the model to the ‘External finances’ section where the ‘Reserve currency flexibility’ is comprised. The adjustments are based on the assessment of the resilience and range of external financing sources, the external debt sustainability and the vulnerability to external shocks. See Fitch (2020l), Section ‘IV. External Finances’.

While Fitch’s methodology specifies that peers refer to both the countries in the quantitative rating category (A in the case of Italy) and those in the official rating category (BBB) as the issuer under assessment, Fitch’s announcements on rating decisions on Italy refer only to the BBB category as peer countries for Italy.

Fitch defines the ‘Net external debt’ as ‘the difference between gross external debt and residents’ debt claims on non-residents’ (see Fitch (2020l), p 26). Based on the IMF definition of the net external debt, this includes only debt instruments, i.e. instruments that require payments of principal and/or interests by the debtor. See IMF (2013).

Source of the figures reported in the text: (i) Eurostat, for the net international investment position; (ii) FitchRatings reports, for the net external debt: Fitch (2020h), p 5; Fitch (2020i); Fitch (2020j), p 4; Fitch (2020f), p 5; Fitch (2020g), p 4; Fitch (2020a), p 2; Fitch (2020d), pp. 2–4; Fitch (2020c). For some countries we considered two different reports since their 2019 NED figures were updated by Fitch in the second half of 2020; instead, the 2019 peer median was available only in the 2020 first-half reports. We used Finland report to retrieve the 2019 NED for France’s AA peer median.

See Fitch (2020l), p 26.

More precisely, the NIIP is the difference between (a) financial assets of residents of an economy that are claims on non-residents and gold bullion held as reserve assets, and (b) liabilities of residents of an economy to non-residents. The difference can be positive or negative and is calculated from the asset perspective (assets minus liabilities). The NED refers only to ‘debt instruments’; it is calculated from the liability perspective as the difference between an economy’s external liabilities and assets in debt instruments (instruments that require payments of principal and/or interests by the debtor). For more details, see: IMF (2009, 2013).

See Moody’s (2019), p 27.

At the end of 2019, France debt-to-GDP ratio was 98.1% against 59.6% for Germany. See Eurostat database (2020).

See S&P (2020a, 2020b).

We conduct the simulation for each of the four agencies as follows. First, in the agency’s published methodology we identify the list of quantitative indicators that are included in the agency’s model. Second, we identify the most suitable data source, if not already specified in the methodology, to get Italy-specific data to be used as input for the model indicators (we use data from the Bank of Italy, BIS, Eurostat, IMF, Italian National Institute of Statistics (Istat), Ministry of Economic and Finance of the Republic of Italy, OECD, World Bank, and World Economic Forum). We then compute the value of each model indicator following the rules specified in the methodology. Next, we attach, based on the indicator value, the corresponding score in the scales/tables published in the agency’s methodology (e.g. a GDP higher than X gets a Y score). Finally, we combine all model scores according to the aggregation rules specified in the methodology (e.g. average, sum, minimum or maximum function). The resulting (simulated) ‘model rating’ only relies on the quantitative variables that are included in each agency’s model (DBRS and Fitch models are fully quantitative, Moody’s and S&P models are quali-quantitative models).

For the four agencies’ model simulations, the difference between our simulation’s and the respective agency’s results depend on differences in the input values for the model indicators. In particular, this happened with those indicators whose model values are also based on forecasts.

See DBRS Morningstar (2020b).

The scoring scale [0, 20] is in turn divided into 11 ‘categories’ which go from ‘very weak’, corresponding to the interval [0, 0.99] to ‘very strong’, corresponding to the range [19.00, 20.00].

DBRS says that the 2-notch negative adjustment comes from a downward revision applied to 3 out of the 6 blocks of indicators in its model: (i) the ‘Economic Structure and performance’ area (from ‘strong/good’ to ‘good’), because of a GDP per capita (33,200 US dollars in 2019) lower than euro area peers’ and the gap on green and digital transition investments; (ii) the ‘Monetary policy and financial stability’ area (from ‘very strong’ to ‘strong’), because of the high NPL stock and the low diversification of some small and medium-sized banks; and (iii) the ‘Political environment’ area (from ‘strong/good’ to ‘good’) for the political uncertainty, as shown by the frequent change in governments and the weak appetite for reforms.

See Fitch (2020b, 2020k).

The weights mentioned in the text refer to Fitch’s methodology in force at the time of our simulation (October 2020). Little changes were applied by Fitch afterwards with the methodology updates.

Fitch uses a linear regression model, hence no scoring scales are available for each indicator to assess the relative position of a sovereign in the regression (e.g. how high is a 3.05 value for the variable GDP-per-capita for Italy, obtained as the product of Italy’s GDP-per-capita, as a percentile rank, times the coefficient of 0.040 for that variable in the regression). Lacking a reference scale, for the purpose of this analysis, the scores assigned by Fitch model to Germany are used as a benchmark to assess the scores assigned by Fitch model to Italy. Data on Germany were sourced from Fitch (2020e), p 2.

The −5.4 point difference between Italy’s and Germany’s quantitative scores derives from: (i) -2.4 points on Pillar 1 (Structural features) which includes, as main indicators, GDP-per-capita and World Bank Governance Indicators; (ii) −2.1 points on Pillar 3 (Public finances), mainly based on the debt-to-GDP ratio; and (iii) around −0.5 points on each of the two remaining pillars in the Fitch model, namely, Pillar 2 (Macroeconomic performance, policies, and prospects), which includes the GDP growth and the inflation indicators, and Pillar 4 (External finances) based on balance of payments indicators.

The Fitch model is structured in such a way that every additional point in the final score (given by the sum of the 4 Pillars’ individual scores) corresponds to 1 notch in the implied credit rating. For example, a final regression score between 6.5 and 7.5 corresponds to a BBB- rating; a final score between 7.5 and 8.5 to a BBB rating, and so on. Therefore, the 5 point difference between Italy and Germany yields a 5 notch difference in the model ratings (A− vs. AA+).

See Fitch (2020l), p 11.

Hence the 3-notch penalisation stems from a downward adjustment to 2 analytical pillar scores: ‘Macroeconomic performance, policy and prospects’ pillar (−1 notch) and the ‘Public finances’ pillar (−2 notches).

The over-performance referred to the following metrics of the model: (i) World Bank Governance indicators, GDP per capita, Share in world GDP, Broad money/GDP ratio (in the ‘Structural features’ pillar) and (ii) Commodity export dependence, Current account balance and Net foreign direct investment, Sovereign net foreign currency debt (in the ‘External finances’ pillar).

Regarding ‘Institutions and governance strength’, Moody’s made this factor mostly qualitative in the methodology review of November 2019, hence it cannot be entirely replicated (formerly, instead, it was based on inflation variables and some World Bank indices). In order to process this factor in our simulation, we take into account the only two quantitative indicators in this area that are also publicly available (namely, the ‘Quality of legislative and executive institutions’ and the ‘Strength of civil society and the judiciary’, both published by the World Bank) which yield an A score. This corresponds to the score assigned by Moody’s in its November 2020 rating review.

Moody’s explains that the choice for the lowest score among those assigned to the four sub-factors depends on the fact that the four sub-factor risks (political, government liquidity, banking sector, and external vulnerability risks) are typically correlated, with the manifestation of one of these risks likely to accelerate the occurrence of other risks’. See Moody’s (2019), p 33.

Moody’s methodology assesses the ‘Banking sector risk’ through two indicators: the ‘Strength of the banking system’ (rating) and the ‘Size of the banking system’ (total assets-to-GDP). These two indicators are combined through a table also reported in Moody’s methodology.

In our simulation, we can replicate only the score on the banking sector risk since this has become the only quantitative sub-factor, within the ‘Susceptibility to event risk’ factor, in Moody’s methodological review of November 2019 which made this factor mostly qualitative. To determine if the banking risk score (BB) is the worst among the four sub-factors, we assess the other three sub-factors as follows: (i) for the political risk, we take into account the only quantitative indicators foreseen by the new methodology for this sub-factor (the World Bank ‘Voice and accountability’ and ‘Political stability’ indices and the country income Gini coefficient) which yields an A score, against a BBB reported by Moody’s in its November 2020 report on Italy; (ii) for the external vulnerability risk, we take into account also in this case the only quantitative indicator specified in the methodology (the current account balance) whose assessment can be ‘scored’ based on what described in the methodology: this leads to an AA score, in line with Moody’s score in November 2020 report; and (iii) for the liquidity risk, we apply Moody’s former, more quantitative methodology which yields an A score, in line with Moody’s report of November 2020.

See Moody’s (2020a), pp. 3–6.

See S&P (2020b).

According to S&P methodology, interest payments at the level of Italy’s, i.e. between 5 and 10% of the revenues, could lead to a better score on the fiscal assessment (between 2 and 5) for a debt-to-GDP ratio lower than 100%; instead, for interests-to-revenues between 5 and 10% coupled with a debt higher than 100% of the GDP, that combination takes the lowest score (6). This means that, at the current and projected debt level, Italy’s assessment in S&P fiscal section could hardly improve.

See S&P (2020a, 2020b).

See S&P (2019).

Data sourced from the Eurostat (for the euro area) and the Office for National Statistics (UK), consultation date: 13 January 2023.

For this specific assessment area, we analyse also Fitch outcome on Italy by taking Germany as a benchmark (as stated above, a benchmark must be used for cross-comparison absent, in Fitch methodology, a scoring scale at indicator level). In Fitch’s model, assuming for Italy the same public finance score assigned by Fitch to Germany, Italy’s model-implied rating would improve by 2.4 notches. The same simulation on the three other agencies—all other indicators being equal—results in a rating differential of 2 notches for S&P and 4 notches for DBRS and Moody’s. Therefore, the government debt indicator has a more significant rating impact in DBRS and Moody’s rating relative to Fitch and S&P.

See Moody’s (2020b), p 2.

Data sourced from Eurostat.