Assessing Next Generation EU

Abstract

The unprecedented fiscal package adopted by the European Council in the summer of 2020—dubbed Next Generation EU—was vital for the recovery of the euro area from the pandemic shock. However, our computations with a stylized macroeconomic model illustrate that an alternative approach, with ex ante risk sharing through the creation of a Eurobond and permanent fiscal capacity at the Centre, would have been at least as powerful, yet it would be more sustainable, automatic and timely.

Lesson from COVID-19. A strong EU: Staying together in a new policy space.

Edited by Luigi Paganetto, Tor Vergata University, Rome.

© Springer Science+Business Media New York 2021.

Annex: A Stylised Model

In an earlier paper (Codogno & van den Noord, 2019) we developed a model to examine how a new set of policy tools—in particular, a safe asset and fiscal capacity at the centre—could improve the resilience of the Eurozone economy to (symmetric or asymmetric) demand and supply shocks. In a subsequent paper (Codogno and Van den Noord, 2020), we extended this analysis to include financial risk-premium shocks stemming from, for example, deterioration of asset quality in periphery banks, political turmoil in the periphery or a fall in global risk appetite. This analysis necessitated a major extension of the model, to include explicit modelling of bond yields, bank lending and public debt dynamics. In the present paper, we have modified this model to capture the impact of the Covid-19 shock and its policy responses.

1.1 A.1 The Real Economy

The aggregate (log-linear) demand equations follow the standard Mundell-Fleming approach adapted to the features of a (closed) monetary union and are perfectly symmetric:

$$ \Big\{{\displaystyle \begin{array}{c}\ {y}^d={\phi}_1l+{\phi}_2\left(f+{f}^{\in }+\ell +{\ell}^{\in}\right)-{\phi}_3\left(\pi -{\pi}^{\ast}\right)-{\phi}_4\left(y-{y}^{\ast}\right)+{\varepsilon}^d\ \\ {}\ {y}^{\ast d}={\phi}_1{l}^{\ast }+{\phi}_2\left({f}^{\ast }+{f}^{\ast \in }+{\ell}^{\ast }+{\ell}^{\ast \in}\right)+{\phi}_3\left(\pi -{\pi}^{\ast}\right)+{\phi}_4\left(y-{y}^{\ast}\right)+{\varepsilon}^{\ast d}\end{array}} $$

(1)

where an asterisk (*) indicates the periphery, and variables without an asterisk refer to the core. Aggregate demand y ^d and y ^∗d is determined by the supply of bank credit l and l ^∗, the fiscal stance—gauged by the primary government deficit f and f ^∗—and cross-border trade. The latter is a function of the inflation differential π − π ^∗ (a proxy for the real exchange rate) and the relative pace of economic growth y − y ^∗. In addition, we include the impact of fiscal policy conducted by the ‘fiscal capacity’, captured by its primary deficit as distributed to each block, denoted as f ^∈ and f ^∗∈ as well as the impact of loans extended from the fiscal capacity to the national sovereigns ℓ and ℓ ^∗. Because these loans are below the line, they do not show up in the fiscal stance either at the centre or at the national level. However, they do have an impact on economic activity. For simplicity, the multipliers for national and supranational fiscal policy are assumed to be the same (i.e. ϕ ₂). Finally, ε ^d and ε ^∗d are demand shocks.

Aggregate supply y ^s and y ^∗s is determined by the inflation ‘surprises’ π − π ^e and π ^∗ − π ^∗e relative to expectations (denoted by the superscript e) alongside exogenous supply shocks ε ^s and ε ^∗s, via an inverted Phillips-curve:

$$ \Big\{{\displaystyle \begin{array}{c}\ {y}^s=\left(\pi -{\pi}^e\right)/\omega +{\varepsilon}^s\\ {}\ {y}^{\ast s}=\left({\pi}^{\ast }-{\pi}^{\ast e}\right)/\omega +{\varepsilon}^{\ast s}\end{array}} $$

(2)

Expected inflation is partly anchored in the official inflation target ${\overline{\pi}}^T$ and is partly backward looking and hence depends on actual domestic inflation:

$$ \Big\{{\displaystyle \begin{array}{c}\ {\pi}^e=\left(1-\eta \right){\overline{\pi}}^T+\eta \pi \\ {}\ {\pi}^{\ast e}=\left(1-{\eta}^{\ast}\right){\overline{\pi}}^T+{\eta}^{\ast }{\pi}^{\ast}\end{array}} $$

(3)

Since all variables are defined as deviations from a steady state in which all shocks are nil, we may assume that ${\overline{\pi}}^T=0$. We allow for the possibility of an asymmetry in the formation of inflation expectations such that η ^∗ ≥ η, which means that potentially there could be greater inflation proneness in the periphery than in the core.

Finally, in equilibrium aggregate demand equals aggregate supply, hence:

$$ \Big\{{\displaystyle \begin{array}{c}\ {y}^s={y}^d=y\\ {}\ {y}^{\ast s}={y}^{\ast d}={y}^{\ast}\end{array}} $$

(4)

The numerical calibration of the parameters is displayed in Table 2. A crucial parameter is the fiscal multiplier ϕ ₂. Mainstream estimates are of the order of 0.5, see for instance Baum et al. (2012) and Barrell et al. (2012), and we adopted this value in our earlier paper (Codogno and Van den Noord, 2020). However, as discussed in more detail in Van den Noord (2020), the magnitude of the fiscal multipliers depends inter alia on the cyclical position of the economy and whether or not a liquidity trap besets monetary policy. Therefore, we have augmented the multiplier to 0.8, crudely based on Batini et al. (2014).

Table 2 Numerical calibration

With regard to the other parameters in the Eqs. (1)–(4) we resort to the calibration in Codogno & Van den Noord (2020). Specifically, for ϕ₁, capturing the impact of bank credit on the real economy (Antoshi et al., 2017) find for 39 European countries a 10% increase in bank credit to boost real GDP by 0.6–1%. However, Cappiello et al. (2010) find a much stronger effect for a panel of Eurozone members, with a 10% increase in credit leading to a 3.2% increase in real GDP. Accordingly, we adopt ϕ₁ = 0.333. Estimates for the parameters that capture cross-border trade, comprising ϕ₃ for absorption and ϕ ₄ for competitiveness, are based on Bayoumi et al. (2011) and ECB (2013), with ϕ ₃ = ϕ ₄ = 0.5.

For the parameter gauging the slope of the Phillips curve ω we again refer to Codogno and Van den Noord (2019), who—based on Ball et al. (2013) and Llaudes (2005)—assumed that ω = 0.25. Finally, Van der Cruijsen and Demertzis (2009) find a strong dependence of inflation expectations on actual inflation in the periphery, but no such relationship in the core. Therefore, we will adopt as our baseline estimate η = 0 and η ^∗ = 0.5.

1.2 A.2 The Financial Sector

A hallmark of the Eurozone predicament is the so-called ‘doom loop’ which refers to tensions in the sovereign debt market prompting a ‘credit crunch’, with the resulting economic slump feeding back into the sustainability of sovereign debt. The main channel through which tensions in sovereign debt markets affect the supply of bank credit is via the cost and the availability of wholesale funding for banks. Financial distress and the associated capital flight from the periphery to core sovereign debt raise the cost and cut the availability of funding for banks in the periphery.

It may be assumed that this source of vulnerability vanishes once Eurobonds, guaranteed by the joint sovereigns, become available. As the national sovereign will lose their eligibility for purchases by the ECB, and Eurobonds would be eligible instead, national sovereigns would become inherently riskier. It, therefore, makes sense that they would also lose their zero-risk weighting. Therefore, it is reasonable to assume that banks agree to swap their sovereign debt portfolio for Eurobonds, on a voluntary basis. As a result, sovereign debt distress, and the associated capital flight from the periphery to the core, no longer matters for the cost or availability of bank funding in the periphery.

Moreover, since all banks have access to the same safe asset, the Eurobond, central bank purchases can be assumed to induce banks to convert the additional (excess) reserves thus created into loans (unlike the current situation where banks keep the excess on their balance sheets as protection against loss of access to wholesale funding). This is known in the literature as the direct bank lending channel of quantitative easing. Evidence of this channel being effective at present in the Eurozone is weak, as banks in practice have been holding on to their excess reserves or used them to pay down external funding or (re-)purchase debt securities instead of providing credit to the economy (see Ryan & Whelan, 2019). However, this may change when banks are induced to hold Eurobonds in lieu of national sovereign bonds. As national sovereign bonds lose their zero-risk weighting, the scope for carry trades diminishes and, with the ‘doom loop’ broken, the need to hold on to excess reserves also diminishes, hence it looks plausible that a direct bank lending channel will open. There is indeed some empirical evidence that a direct bank lending channel is effective in cases where banks have access to a (national) safe bond, see Paludkiewicz (2018) for Germany, (Joyce & Salto, 2014) for the UK and Kandrac and Schlusche (2017) for the US.⁵

These notions are embedded in the following stylised (log-linear) equations for bank credit measured as a percentage of nominal output, in which the periphery-core yield spread r ^∗ − r is included as a gauge of sovereign debt distress:

$$ \Big\{{\displaystyle \begin{array}{c}l-\left(y+\pi \right)=-{\xi}_1i+s{\xi}_2q-\left(1-s\right){\xi}_3\left(r-{r}^{\ast}\right)+\lambda \\ {}\ {l}^{\ast }-\left({y}^{\ast }+{\pi}^{\ast}\right)=-{\xi}_1^{\ast }i+s{\xi}_2^{\ast }q-\left(1-s\right){\xi}_3^{\ast}\left({r}^{\ast }-r\right)+{\lambda}^{\ast }\ \end{array}} $$

(5)

and where λ and λ ^∗ are exogenous shocks to the respective banking systems (credit crunch or credit boon). Moreover, q denotes the purchases of sovereign bonds by the ECB as a percentage of GDP, and i is the ECB’s main policy rate (for simplicity we abstract from the distinction between the deposit and the repurchase rate, and s is a dummy variable which takes the value 1 if a Eurobond is created and which is nil otherwise. We expect that ${\xi}_1^{\ast}\ge {\xi}_1$, ${\xi}_2^{\ast}\ge {\xi}_2$ and ${\xi}_3^{\ast}\ge {\xi}_3$, so generally speaking the sensitivity of bank lending to monetary policy and financial market distress would be larger in the periphery than in the core. Note also that there is an asymmetry in the sense that the adverse effect of the yield spread on lending in the periphery has the opposite sign of the safe-haven effect on lending in the core, and that both tend to widen the differential.

This takes us to the determinants of the sovereign yield spread of the Eurozone periphery against the core r ^∗ − r. There is burgeoning literature on the sovereign yield spread in the Eurozone, which is usually assumed to be driven by country-specific liquidity risk, country-specific default risk and the risk appetite of global investors (see, for instance, Codogno et al., 2003). The ratio of sovereign debt to GDP (alongside the fiscal deficit feeding into the debt ratio) is usually considered to be the main driver of country-specific default risk. As several studies have shown, the relationship between debt and spread can be strongly non-linear and dependent on global risk sentiment. With the outbreak of the global financial crisis, the perception of higher sovereign default risks produced a sharp increase in yield spreads, and even more so in countries whose initial debt ratio was comparatively high.

By contrast, as indicated inter alia by De Grauwe and Ji (2012), in developed economies with a federal/central government that issues debt in its ‘own’ currency, federal sovereign yields tend to incorporate liquidity and exchange rate risk premiums, but not a default risk premium. A Eurobond, issued by an appointed fiscal capacity with full democratic legitimacy, and which enjoys a joint guarantee by the national sovereigns, may be assumed to fit this description broadly. However, once a Eurobond exists, the national sovereigns would become more akin to state and local government debt in federal states, i.e. would still carry default risk premia (see Schuknecht et al., 2009). In fact, due to the joint guarantee (and assuming this guarantee is credible), national sovereign debt would become inherently riskier than at present, with their yields incorporating risk premia not only for national but also for supra-national public debt.

These features are reflected in the following set of equations for national and supranational yields:

$$ \Big\{{\displaystyle \begin{array}{c}\ r=s{r}^{\in }+\left(1-s\right)\left({\vartheta}_1i-{\vartheta}_2q\right)+{\vartheta}_3\left(b+s{b}^{\in}\right)+\rho\ \\ {}\ {r}^{\ast }=s{r}^{\in }+\left(1-s\right)\left({\vartheta}_1^{\ast }i-{\vartheta}_2^{\ast }q\right)+{\vartheta}_3^{\ast}\left({b}^{\ast }+s{b}^{\in}\right)+{\rho}^{\ast}\\ {}\ {r}^{\in }=\left(1-s\right)\frac{1}{2}\left(r+{r}^{\ast}\right)+s\left({\sigma}_1i-{\sigma}_2q+{\rho}^{\in}\right)\ \end{array}} $$

(6)

where r, r ^∗ and r ^∈ are the yields on core, periphery and supranational sovereign debt and b, b ^∗ and b ^∈ denote the corresponding sovereign debt as a per cent of GDP. The variables ρ, ρ ^∗ and ρ ^∈ are exogenous risk premium shocks. Moreover, q again denotes the purchases of sovereign bonds (regardless of the issuer) by the ECB, as a percentage of GDP, and i is again the ECB’s primary policy rate. We expect ${\vartheta}_1^{\ast}\ge {\vartheta}_1$, ${\vartheta}_2^{\ast}\ge {\vartheta}_2$, ${\vartheta}_3^{\ast}\ge {\vartheta}_3$, so generally speaking periphery yields are the most sensitive to developments in sovereign debt and monetary policy. Let us recall that all variables (except for the dummy s) are defined in terms of deviations from a baseline in which all shock variables are nil. The idea is to leave these equations unchanged on the assumption that the yield of Eurobonds would follow the same pattern as ESM bonds, i.e. a weighted average of the underlying national sovereign bonds.

The numerical assumptions for the system of Eqs. (5) and (6) are again as much as possible based on the mainstream literature (see Table 1). For bank lending, Albertazzi et al. (2012) find for Italy (which we take to represent the periphery) an adverse effect of a 100 bps increase in the spread r ^∗ − r of the order of 3.5% for loans to NFCs and 6.0% for household loans. Given the relative weights of NFC and household loans, this implies that approximately ${\xi}_3^{\ast }=4.5$. Based on the same study we adopt ${\xi}_1^{\ast }=3.0$ for the impact of the policy rate on bank credit, although estimates were based on the peak of the government bond crisis and by now the sensitivity has decreased significantly. In the core, we assume the impact of the spread r ^∗ − r to be nil such that ξ ₃ = 0, as suggested by Altavilla et al. (2016). For the impact of quantitative easing on bank lending we adopt ${\xi}_2^{\ast }=0.25$, i.e. for every euro liquidity created on banks’ balance sheets in the periphery through asset purchases, one-quarter is converted into bank loans. This is in line with findings for the United Kingdom reported by Joyce and Salto (2014). Our baseline assumption for the effectiveness of quantitative easing in the core is smaller than in the periphery, with ξ ₂ = 0.125, to reflect the smaller holdings of sovereigns on banks’ balance sheets.

The numerical calibration of the yield equations is based on De Santis (2016). Accordingly, we adopt for the impacts on yields of the policy rate ${\vartheta}_1={\vartheta}_2^{\ast }=0.5$, with the impact thus less than proportional to reflect that tighter monetary policy now gets countries loser monetary policy later, so bond yields will not increase as much as policy rates. With regard to the impact of quantitative easing on sovereign yields we adopt ϑ ₂ = 0.05 and ${\vartheta}_2^{\ast }=0.1$. This implies that for every 1% of GDP equivalent of asset purchases by the ECB, yields would drop by 5 bps in the core and by 10 bps in the periphery. Note that total asset purchases by the ECB to date have roughly amounted to around 25% of GDP, which according to the above estimates would have slashed yields by 100 bps in the core and 250 bps in the periphery. Finally, based on the same study, we adopt for the impact of the public debt ratio on the sovereign yields ϑ ₃ = 0.23 and ${\vartheta}_3^{\ast }=0.26$.

Obviously, we do not know how the yield on Eurobonds will behave in response to monetary policy. Therefore, we will assume the impact of ECB asset purchases on the Eurobond yields to average that on the national sovereign yields when s = 0, so σ ₁ = 0.5 and σ ₂ = 0.075.

1.3 A.3 The Government Sector

The usual debt dynamics identities capture the evolution of the debt ratio to output at the national and supranational levels. We also allow for discretionary fiscal spending (grants) and loans at the centre to differ between the core and the periphery:

$$ \Big\{{\displaystyle \begin{array}{c}b={b}_0\left(\chi r-y-\pi \right)+f+\ell +{\ell}^{\in }\ \\ {}{b}^{\ast }={b}_0^{\ast}\left(\chi {r}^{\ast }-{y}^{\ast }-{\pi}^{\ast}\right)+{f}^{\ast }+{\ell}^{\ast }+{\ell}^{\ast \in }\ \\ {}{b}^{\epsilon }=s{b}_0^{\in}\left(\chi {r}^{\in }-\overline{y}-\overline{\pi}\right)+\frac{1}{2}\left({f}^{\in }+{\ell}^{\in }+{f}^{\ast \in }+{\ell}^{\ast \in}\right)\end{array}}\vspace*{-6pt} $$

(7)

$$ \Big\{{\displaystyle \begin{array}{c}f=-\left(\tau - s\theta \right)y+g,{f}^{\in }=- s\theta y+{g}^{\in }\ \\ {}\ {f}^{\ast }=-\left(\tau - s\theta \right){y}^{\ast }+{g}^{\ast },{f}^{\in \ast }=- s\theta {y}^{\ast }+{g}^{\ast \in}\end{array}} $$

(8)

where $\overline{\pi}=\frac{1}{2}\pi +\frac{1}{2}{\pi}^{\ast }$ and $\overline{y}=\frac{1}{2}y+\frac{1}{2}{y}^{\ast }$ and where g, g ^∗, g ^∈ and g ^∗∈ denote the discretionary component of the respective deficits, ℓ and ℓ ^∗ are loans from the national governments to the private sector, ℓ ^∈ and ℓ ^∗∈ are loans from the centre national governments, and τ corresponds to the usual ‘semi-elasticity’ of the fiscal deficit with respect to output. In this specification, sθ takes a positive value when a supra-national fiscal capacity is created, and certain tax or spending programmes are reallocated to it, and nil otherwise. The primary deficit at the central level is simply the average${\overline{f}}^{\in }=\frac{1}{2}{f}^{\in }+\frac{1}{2}{f}^{\in \ast }$.

Let us recall that f, f ^∗, f ^∈ and f ^∗∈ denote the respective primary deficits as a ratio to output that enters the system of aggregate demand Eq. (1) and that b ₀, ${b}_0^{\ast }$ and ${sb}_0^{\in }$ are the respective ‘initial’ debt ratios, whereby we mean the prevailing debt ratios if none of the potential demand, supply or financial shocks occur (i.e. ε ^d = ε ^∗d = ε ^s = ε ^∗s = λ = λ ^∗ = ρ = ρ ^∗ = 0). As before, if s = 0 no Eurobonds are created, so b ^ϵ = 0. However, if s = 1, the debt ratio would change in response to variations in the relevant yields, economic growth and inflation alongside the conduct of fiscal policy at the centre. We make a simplifying assumption that a fraction χ of the changes in yields feed through in the implicit debt servicing cost, depending on the percentage of the total stock of debt that comes due each year. In the model simulations, it is assumed that χ = 0.2.

The primary fiscal deficits f, f ^∗, f ^∈ and f ^∗∈ are partly endogenous on account of ‘automatic stabilisers’ (e.g. variations in tax proceeds or social security outlays as a function of cyclical economic activity), so they comprise induced and discretionary components. For the numerical calibration of the automatic stabilisation effect, we refer to Van den Noord (2000) and Girouard and André (2005), which implies that τ = 0.5. Furthermore, we assume that b ₀ = 50%, ${b}_0^{\ast }=130\%$ and ${b}_0^{\in }=40\%$. This roughly corresponds to, respectively, the public debt to GDP ratios in Germany and Italy and the amount of Eurobonds that approximately needs to be issued to cover the purchases of national sovereigns on the balance sheets of the ECB and the banks as well as any additional purchases in the market needed to secure consistency with the capital key. As concerns the parameter θ we refer to Van den Noord (2020), who assumes that half of the automatic stabilisation effect would accrue to the centre, so if τ = 0.5 then θ = 0.25.

1.4 A.4 Shocks and Changes in Policy Variables

As discussed in the main text, three scenarios are computed. The exogenous changes assumed in each of these three scenarios are reported in Table 3 below. Specifically,

• 1. The dummy s takes a value 0 in Scenarios I and II and 1 in Scenario III.

• 2. In all three scenarios the same set of demand shocks ε ^d and ε ^∗d and supply shocks ε ^s and ε ^∗s are assumed as well as the same change in the policy rate i. Also, in all three scenarios the same exogenous risk premium shocks to sovereign yields ρ and ρ ^∗ are incorporated to reflect a flight to safety effect on core yields and an offsetting (neutralising) effect of the ESM emergency facility on periphery yields.

• 3. The domestic fiscal shocks g are identical across the three scenarios except from a reduction in Scenario II to reflect the impact of grants from the centre used to replace deficit funding of domestic spending. The same holds for the domestic fiscal shock in the core g ^∗.

• 4. In Scenario I the increase in central public spending g ^∈ and g ^∗∈ is modest, reflecting the first batch of EU programmes in the spring such as SURE. The sharp increases in these fiscal variables (especially in the periphery) in Scenario II reflect the grants provided under New Generation EU. The same holds for the increase in loans from the centre ℓ ^∈ and ℓ ^∗∈ in Scenario II relative to Scenario I.

• 5. In Scenario III the aggregate amounts of grants and loans from the centre are the same in Scenario II, but the distribution across the core and symmetry is now symmetric, meaning that g ^∈ = g ^∗∈ and ℓ ^∈ = ℓ ^∗∈.

Table 3 Shocks and changes in policy variables