3.1. A historic investment and reform package

NGEU aims to repair the immediate economic and social damage brought about by the COVID-19 pandemic and make Europe fit for current and future challenges. Designed in response to the COVID-19 pandemic, it is a temporary instrument to boost the EU’s long-term budget (the multiannual financial framework, 2021–2027).

3.2. A stylized composition and allocation

Modeling NGEU requires several simplifying assumptions. Firstly, we broadly partition the total package into grant and loan instruments, summarized in Table 1, totaling around 4% of EU GDP. NGEU was designed to favor lower-income and vulnerable countries and those particularly hard-hit by the pandemic. A large share of the NGEU package boosts public investment through grants. The allocation of grants across Member States is based on the current Recovery and Resilience Facility (RRF) maximum grant allocation. In total, the simulations assume that EUR 396 bn (in 2019 prices or EUR 421 bn in current prices) will be provided as grants.Footnote 9

On loans, we assume that seven Member States request a total of 166 bn EUR in RRF loans, based on information as of June 2021.Footnote 10 Thus, the simulated loan share is significantly below the loan facility’s cap at 6.8% of the gross national income of the applicant Member State. Note, however, that the loan amount is expected to increase, because several Member States have indicated that they intend to ask for loans at a later stage. Graph 1 provides an overview.

Graph 1.

Overview of assumed allocation. Note: This graph reports the assumed grant and loan allocation used in the simulations. Note that this is a highly stylized representation for modeling purposes only; actual sums financed from NGEU are bound to differ. Grant instruments include RRF grants and additional resources such as ReactEU and the Just Transition Fund. Two-letter country codes follow EU conventions (https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Country_codes ).

Table 1.

Apportioning across NGEU instruments (for modeling purposes only)

Note: This table reports the assumed grant and loan composition used in the simulations in 2019 prices. Note that this is a highly stylized representation for modeling purposes only; actual sums financed from NGEU are bound to differ. Grant instruments include RRF grants and additional resources such as ReactEU and the Just Transition Fund, which share economic characteristics but follow a different allocation key in the actual implementation, which the simulations only partly reflect.

3.3. Financing assumptions

We distinguish between assumptions on grant financing and assumptions on loan financing. For grant financing, the simulations assume that the EU debt is long term (average maturity of around 16 years). The repayment starts at the end of the current multiannual financial framework in 2027 and ends in 2058, following a linear schedule. It is further assumed that all Member States contribute to the EU budget according to their current GDP shares, subtracting from future changes in the GNI shares or new own EU resources. Domestic lump-sum taxes finance these contributions.Footnote 11

The RRF loan repayment by the receiving Member States begins in 2031 and ends in 2050 (following a linear schedule). Interest rates for highly indebted countries are at a more favorable, lower interest rate. Loan repayments by the Member States are also financed via domestic lump-sum taxation.Footnote 12

3.4. Further simplifying assumptions

We make three additional simplifying assumptions. Firstly, the simulations assume an even allocation across the years of NGEU’s active operation. We consider a 6-year profile (i.e. 16.67% each year from 2021 to 2026). Online Appendices F and I report the results for a fast scenario, featuring an even allocation across 4 years (2021–2024). The assumed profile is the same for all NGEU components and for all Member States.

Secondly, the simulations assume that the overall NGEU allocation is spent as productive public investment. In national accounts terms, spending on education and training may be classified as consumption, but for modeling purposes, we consider it as productive spending (see next section).Footnote 13

Finally, the simulations assume that Member States use 100% of EU grants for additional public investment, while it is assumed that EU loans are 50% additional. Since the other half of loans finances general government spending, which would take place anyway (and thereby frees up resources), the impact on the national debt is also 50%.Footnote 14

4.1. Fiscal policy

4.1.1. Public investment: productivity effects

A central assumption in our study is that public investment is productivity-enhancing, a notion broadly supported by the empirical literature [see Bom and Ligthart (Reference Bom and Ligthart2014); Ramey (Reference Ramey2020)], despite identification challenges. Formally, we capture productivity effects by including public capital in the private sector’s production process. Higher public capital then increases output for given inputs (private capital and labor). Following Baxter and King (Reference Baxter and King1993), we can write a simplified representation of the private sector production function as:

(1) \begin{equation}Y_{t}=N_{t}^{\alpha }K_{t}^{1-\alpha }\left(K_{t}^{G}\right)^{{\alpha _{G}}},\end{equation}

where $Y_{t}, K_{t}, N_{t}$ , $\alpha$ , and $K_{t}^{G}$ denote output, private capital, labor, the labor share, and effective public capital, respectively.Footnote 16 The output elasticity of public capital, $\alpha _{G}\geq 0$ , drives the medium- and long-run GDP effects in our simulations. We discuss its calibration and sensitivity below.

Besides its supply-side effects, public investment directly enters GDP in the national account expenditure items. Therefore, other things being equal (absent crowding-out effects), higher investment demand drives up output independently of our productivity assumptions. Hence, public investment in the model increases aggregate demand in the short run and aggregate supply in the medium and long run.

4.1.2. Public investment: time-to-build and time-to-spend

Public investment often faces delays in both implementation and construction. For example, projects need to be contracted.Footnote 17 New infrastructure projects take time before benefiting their users (e.g. building highways or bridges). Standard approaches [e.g. the seminal contribution of Baxter and King (Reference Baxter and King1993)] often set these issues aside. By contrast, we extend the standard model with time-to-build and time-to-spend delays, along the lines of Leeper et al. (Reference Leeper, Walker and Yang2010).Footnote 18

These features have two main implications. Firstly, government investment is not immediately productive, reflecting time-to-build lags. Thus, in contrast to the standard model, government investment does not translate directly into productivity-enhancing public capital. Instead, with the time-to-build delay, the positive supply-side effects materialize later, reducing the short-run multiplier. Nonetheless, they remain persistent as public capital depreciates only slowly. Formally, effective public capital (entering private sector production) follows the law of motion:

(2) \begin{equation}K_{t}^{G}=\left(1-\delta ^{g}\right)\!K_{t-1}^{G}+A_{t-N}^{IG},\end{equation}

where $A_{t-N}^{IG}$ denotes authorized investment at time $t-N$ and $\delta ^{g}$ the depreciation rate of public capital. We model NGEU as shocks to authorized investment.

Secondly, the extended model reflects the fact that not all projects are “shovel-ready” due to planning and contracting time. Such time-to-spend delays (Ramey, Reference Ramey2020) induce lags between authorized investment (appropriations) and implemented government investment following:

(3) \begin{equation}IG_{t}=\sum _{n=0}^{N}\psi _{n}A_{t-n}^{IG},\end{equation}

where the parameters $\psi _{n}$ , with $n\in \{0,\ldots N\}$ , govern the speed of implementation.Footnote 19 With this feature, authorized investment only gradually leads to higher (public) investment demand. Thus, unlike in the standard model, the positive, direct demand-side effects also do not unfold immediately.Footnote 20

4.1.3. Government budget

Real government debt ( $B_{t}^{G}$ ) evolves according to:

(4) \begin{equation}B_{t}^{G}=\left(1+r_{t-1}^{g}\right)\!B_{t-1}^{G}+Exp_{t}-R_{t}^{G}-GR_{t}^{EU}+CO_{t}^{EU}+\omega ^{EU} r_{t-1}^{g}B_{t-1}^{G,EU},\end{equation}

where $Exp_{t}$ and $R_{t}^{G}$ summarize the government’s expenditure and revenues, respectively.Footnote 21 The real interest on bonds ( $r_{t-1}^{g}$ ) accounts for a gradual pass-through of policy rates into effective government financing costs associated with the maturity structure of government debt. In the long run, lump-sum taxes stabilize the debt-to-GDP ratio. Receiving a grant ( $GR_{t}^{EU}$ ) decreases government debt.Footnote 22 By contrast, loans increase debt. These back-to-back loans will be repaid gradually by the beneficiary Member States. We assume that lump-sum contributions ( $CO_{t}^{EU}$ ) finance the EU budget in the long run. $B_{t}^{G}$ comprises RRF-specific loans and “traditional” government debt. A fiscal expansion financed via RRF loans avoids a widening of interest-rate spreads. By contrast, in a scenario without the favorable RRF loan rate, a fiscal expansion would imply an increase in the government-bond rate. The term $r_{t-1}^{g}B_{t-1}^{G,\text{EU}}$ captures contributions to interest-rate payments of EU debt, weighted by the country’s GDP share in the EU, denoted $\omega ^{\textrm{EU}}$ . The EU budget aggregates the EU debt issued to finance grants and loans.

4.2. Monetary policy and the zero lower bound

The monetary policy reaction and the exchange rate are important transmission channels of NGEU. Monetary policy in each currency area follows a Taylor rule with a smooth response to inflation and the output gap. Euro-area countries follow a common monetary policy, while we assume an exchange-rate peg (allowing for a small bandwidth) for countries participating in the ERM-II. The remaining Member States implement their independent national monetary policy with a floating exchange rate. We assume that monetary policy is accommodative for six quarters in response to the investment stimulus.Footnote 23 Below, we also simulate the model without this assumption to gauge the role of monetary accommodation.

4.3. Household heterogeneity and sticky wages

A rapidly growing literature has emphasized the role of household heterogeneity as important for the transmission of macroeconomic policy (e.g. Kaplan et al. Reference Kaplan, Benjamin Moll and Giovanni L.Violante2018). Given the richness of the multicountry setup, we follow the literature on fiscal policy and include a less involved model of household heterogeneity which nonetheless captures key insights. This formulation distinguishes (optimizing) Ricardian and liquidity-constrained households (rule-of-thumb consumers). The latter households do not participate in financial markets and consume their entire disposable income in every period. Together with imperfect labor and goods markets, this feature implies a higher sensitivity of consumption to income, generating Keynesian effects of fiscal stimulus, in line with empirical evidence [e.g. Galí et al. (Reference Galí, López-Salido and Vallés2007)].

4.4. International linkages

At the heart of our spillover analysis is a rich trade structure linking individual economies. While adding complexity, this approach captures the transmission of NGEU in the highly integrated EU economy. We add to the existing literature by combining these elements in a large-scale, multicountry model with 28 regions and detailed, cross-border production linkages. We sketch the main features here, relegating mathematical details to Online Appendix B. The model distinguishes T and NT goods and services and explicitly features imported intermediate inputs. The latter capture cross-border value chains and have significant implications for spillovers.Footnote 24 We also distinguish a T and a NT sector to capture realistic, real exchange-rate dynamics in response to the public investment shock.

We formalize this setup as follows. Let $Z=C+G+IG$ denote private consumption ( $C$ ), government consumption ( $G$ ), and public investment ( $IG$ ).Footnote 25 Preferences of private households and the government for T and NT goods follow CES functions as follows:

(5) \begin{equation}Z_{t}=\begin{array}{c} \left(\left(1-s^{T}\right)^{1/\sigma _{ tnt}}\!\left(Z_{t}^{NT}\right)^{\frac{\sigma _{tnt}-1}{\sigma _{ tnt}}}+\left(s^{T}\right)^{1/\sigma _{tnt}}\!\left(Z_{t}^{T}\right)^{\frac{\sigma _{tnt}-1}{\sigma _{tnt}}}\right) \end{array}^{{\sigma _{tnt}}/({\sigma _{tnt}}-1)},\end{equation}

where $Z_{t}^{NT}$ is an index of domestic demand across NT varieties, and $Z_{t}^{T}$ is a bundle of domestically produced ( $Z_{t}^{T,D}$ ) and imported T goods ( $Z_{t}^{T,M}$ ):

(6) \begin{equation}Z_{t}^{T}=\left(\left(1-s_{m}\right)^{1/\sigma _{x}}\!\left(Z_{t}^{T,D}\right)^{\frac{\sigma _{x}-1}{\sigma _{x}}}+{s_{m}}^{1/\sigma _{x}}\!\left(Z_{t}^{T,M}\right)^{\frac{\sigma _{x}-1}{\sigma _{x}}}\right)^{\frac{\sigma _{x}}{\sigma _{x}-1}} .\end{equation}

In equations (5) and (6), $\sigma _{tnt}$ and $\sigma _{x}$ denote the elasticity of substitution between T and NT goods and between the bundles of domestically produced versus imported T goods, respectively. The steady-state shares of T goods in $Z_{t}$ and imports in $Z_{t}^{T}$ are $s^{T}$ and $s_{m}$ , respectively. The intermediate inputs in the T and NT sectors are also composites of T and NT analogously to equations (5) and (6), with T being domestically produced or imported. Total imports ( $M_{t}$ ) are a CES bundle of bilateral imports from all foreign regions f:

(7) \begin{equation}M_{t}=\left(\sum _{f}\left(s^{\,f}\right)^{\frac{1}{\sigma _{1}}}M_{t}^{{f}^{\frac{\sigma _{1}-1}{\sigma _{1}}}}\right)^{\frac{\sigma _{1}}{\sigma _{1}-1}},\end{equation}

where $\sigma _{1}$ is the elasticity of substitution between imports of different origins $M_{t}^{f}$ and $s^{f}$ is the steady-state share of region f in the domestic economy’s imports. Thus, the production structure in each region is linked to intermediate imports from other regions, a key feature in the transmission of fiscal policy. Graphs B.1–B.4 in the online appendix show the detailed (nested) structures for production, final demand, and intermediate demand.

4.5. Real frictions

As typically assumed in larger DSGE models, goods production in our setup also features variable capacity utilization, capital adjustment costs, and labor adjustment costs. In line with estimated model versions [Ratto et al. (Reference Ratto, Roeger and in ’t Veld2009); Albonico et al. (Reference Albonico, Calès, Cardani, Croitorov, Di Dio, Ferroni, Giovannini, Hohberger, Pataracchia, Pericoli, Pfeiffer, Raciborski, Ratto, Roeger and Vogel2019)], these model features help to capture the economy’s dynamic behavior.

4.6. Calibration strategy

5.1. Simulation setup to quantify spillover

We quantify spillover effects in three steps. Firstly, we simulate all NGEU investment plans jointly, that is, the actual synchronized plan. Secondly, we run 27 stand-alone simulations based on the (counterfactual) unilateral plans, that is, assuming that only one Member State implements the investment plan at a time. In a final step, we calculate the fiscal GDP spillover for each country as the difference between the GDP effects in the first and the second simulation.

5.2. EU-wide results: large spillovers

NGEU’s macroeconomic spillovers are significant. Graph 2 shows that the EU-wide GDP effects are around one-third larger when explicitly accounting for spillover effects. Real GDP in the EU-27 is estimated to be more than 1.2% higher in 2026 compared to a no-policy change baseline (blue lines and right panel). Despite the amplification during the zero-lower-bound period, the time-to-build and time-to-spend delays imply that the output effects unfold gradually. The peak effects materialize at the end of 2026 (due to spending delays, a fraction of public investment continues in the following year). A faster implementation implies larger peak-GDP effects, while the long-term effect (2035) is nearly identical.

Graph 2.

Main simulation results: The role of spillovers on EU-27 GDP. Note: This graph reports the level of real GDP in percentage deviation from a no-policy change (no-NGEU) baseline. Blue lines show simulation results from a simultaneous investment stimulus (NGEU, 6-year implementation). Orange dashed lines display a synthetic EU-wide GDP (weighted average) obtained by aggregating 27 stand-alone simulations with unilateral stimulus in each country (6-year implementation). All values are yearly averages of the quarterly series.

By contrast, ignoring positive spillovers reduces the macro impact significantly. We can construct a synthetic EU GDP by aggregating the 27 stand-alone simulations of the unilateral plans (orange dashed lines). Excluding spillover effects in this way, we find GDP effects of around 0.8%–0.9%. Thus, simply aggregating the individual effects of the national RRPs would substantially underestimate the overall growth effects of NGEU.

The level of real GDP remains persistently high even after the disbursement periods: the higher stock of public capital raises the marginal productivity of private production factors under the assumption of productive government investment. As a result of this productivity boost, sizeable medium-run gains in real wages accompany the rise in real GDP.

5.3. Inspecting the mechanism

5.3.2. Spillover

Two main channels contribute to the large spillovers: direct-trade effects and exchange-rate effects. Firstly, the increase in domestic activity and import demand is the most direct source of positive GDP spillovers. With trade in intermediate inputs, the positive spillover to the import demand and foreign GDP relates to imports of final goods and intermediate inputs into domestic production. This spillover effect will particularly benefit export-intensive countries because of the rising demand from trading partners. In the short run, the lower relative price of foreign-produced goods also shifts demand partially to foreign exports. Importantly, third countries transmit these spillover effects further: the additional production of export goods requires imported intermediate goods from other sources. These channels are more powerful for higher fiscal multipliers because of the stronger output response. Furthermore, in the medium run, the positive supply-side effects of the government investment shock lead to a depreciation of the real effective exchange rate, that is, European export prices increase less than the prices of foreign goods and services. EU exports therefore increase.

Secondly, there is an additional (nominal) exchange-rate effect. At the effective lower bound, with accommodative monetary policy, relatively lower real interest rates imply (other things being equal) a euro depreciation, that is, prices of domestic goods increase less than prices of foreign goods. The exchange-rate movement then supports exports. Absent exchange-rate policies, this positive short-run spillover effect is lacking or even reversed for non-euro-area Member States, which do not participate in ERM-II.

Graph 3.

Inspecting the mechanism: EU-27 GDP. Note: This graph reports the level of real GDP in percentage deviation from a no-policy change baseline based on a 6-year profile. Blue lines show simulation results from the baseline model (NGEU). Yellow (dashed-dotted) lines display simulations without an effective lower-bound (ZLB) constraint. Orange (dashed) lines display a low-productivity scenario, setting the output elasticity of public capital ( $\alpha ^{G}$ ) to 0.05.

Graph 4.

Macroeconomic transmission. Note: This graph reports the government balance (in % of GDP) and inflation in percentage points deviation from a no-policy change baseline. All other variables are expressed in percentage deviation from baseline. All results refer to simulation results from the baseline model (NGEU), assuming a 6-year implementation. Orange lines (dashed) display the low-productivity scenario.

5.4. Cumulative multipliers and long-run effects

Turning to the medium and long run, we find that cumulative multipliers can be sizeable when government capital is productive. We define cumulative multipliers as the ratio of the additional GDP to the fiscal stimulus.Footnote 32 Compared to government consumption, public investment can achieve a sizeable long-run effect by raising long-term productivity. The cumulative multipliers, reported in Table 2 and Graph 5, are in the lower range of the multipliers reported in Ramey (Reference Ramey2020, p. 54). For a closed economy (based on a US calibration and a New Keynesian setting), she finds undiscounted long-run multipliers for government investment between 2.9 and 9.8, depending on the assumed productivity of government investment and the initial stock of public capital.Footnote 33 The counterfactual stand-alone simulations also show that excluding spillovers considerably reduces the long-run multiplier.

Graph 5 illustrates that the dynamic, medium-run, and long-run GDP effects depend crucially on the assumed output elasticity of public capital. To see this, we show the cumulative multipliers for the baseline model (blue bars) and the low-productivity scenario (red bars), where the output elasticity of public capital is significantly lower.Footnote 34 For the more optimistic calibrations, the level of real GDP remains substantially higher even after the implementation period: the higher stock of public capital persistently raises the marginal productivity of private production factors. While sizeable growth effects remain even under more pessimistic assumptions, the changes across assumptions are noteworthy.

Table 2.

Illustrative comparison of long-run multipliers (EU-wide)

Note: This table compares the long-run multipliers of our study with those reported in Ramey (Reference Ramey2020, p. 54, New Keynesian model). Our high- and low-productivity settings correspond to $\alpha ^{G}=0.12$ and $\alpha ^{G}=0.05$ , respectively. In the last row, we apply the same discount factor as Ramey (4% p.a.). Graph 5 shows dynamic results for a lower discount rate (closer to currently observed real interest rates). Multipliers correspond to the ratio of the integrals of the GDP gains and the NGEU funds. We report results including spillovers and a counterfactual synthetic EU aggregate based on the unilateral simulations (no spillovers), as outlined in Section 5.1.

Graph 5.

Dynamic cumulative multipliers (including spillover effects). Note: This graph reports the cumulative GDP multipliers. The multipliers are defined as the ratio of the integrals of the impulse responses of output and the NGEU funds. Blue bars show simulation results from the baseline model (NGEU). Red bars display simulations for a low-productivity scenario. All simulations include spillover effects and refer to a 6-year profile. The left panel shows the undiscounted multiplier, while the middle and right panel display discounted multipliers using a real interest rate of 1.5% (p.a.) and 4% [p.a., as in Ramey (Reference Ramey2020)], respectively.

5.5. A closer look at country-specific effects

Even Member States that receive a small allocation of the fund benefit significantly from spillovers from other countries’ RRPs. Indeed, in particular for open economies with smaller grant allocations, spillover effects account for the bulk of the GDP impact. In some cases of very small allocations, for example, Luxembourg and Ireland, positive spillovers explain most of the total impact.

Graph 6 displays the peak-GDP effect for all countries, showing that NGEU strongly supports convergence. Given the allocation key, the Member States with below-average levels of GDP per capita are estimated to experience the largest boost to GDP levels. For a 6-year stimulus and a high-productivity calibration, the increase in output reaches more than 3.2% in Greece; around 3% in Bulgaria, Croatia, and Romania; and around 2.5% in Italy and Portugal. For these countries, the GDP spillover is smaller because their trade partners receive smaller allocations and the economies tend to be less integrated in production chains and trade. The peak effects are larger for the 4-year NGEU scenario but smaller for the low-productivity scenario (see Online Appendix I).

Graph 6.

Effects across countries (6-year profile, high productivity). Note: This graph reports the level of real GDP in 2026 expressed in percentage deviation from a no-policy change baseline and for a 6-year profile (even allocation across 2021 until 2026 for all Member States). Blue bars show simulation results from a simultaneous investment stimulus (NGEU). Spillover (orange) is defined as the difference of the coordinated simultaneous NGEU stimulus in all Member States and the stand-alone simulations of the national plans. Two-letter country codes follow EU conventions (https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Country_codes ).

Especially for Member States outside the euro area, the monetary policy reaction matters for the short-run spillovers. There can be a negative short-run spillover for those countries due to the national currency appreciation (although the total GDP effects remain positive). However, this exchange-rate effect is temporary and becomes positive in the second or third year of NGEU. Additional simulations (reported in Online Appendix H) show that if monetary policy in these Member States partially targets the euro exchange rate, NGEU spillover becomes positive immediately.

Table 3 shows spillover effects in the peak year, that is, the sixth year in the 6-year profile, for all (counterfactual) unilateral plans and for all Member States. This table highlights the importance of the relative NGEU allocations and bilateral trade linkages for spillovers (see also the trade matrix in Online Appendix C). For example, the increase in investment in Belgium, which receives a relatively small allocation of NGEU funds, boosts GDP by 0.37% in Belgium and has small spillover effects, the largest of which is in Luxembourg (where spillovers from Belgium increase GDP by 0.03%). The role of bilateral trade linkages becomes clearer when focusing on the more significant recipients of NGEU funds. Greece receives a relatively large share of NGEU, which boosts Greek GDP by 2.92%, but spillovers are relatively modest (the largest of these is in Cyprus, where spillovers from Greece increase GDP by 0.09%). The spillover effects of Spanish public investment are largest for Portugal (0.15%), given the close trade linkages between the two countries. But overall, the largest spillover effects come from Italy, a large country, receiving a major share of NGEU funds. Note that these spillovers are often larger than what bilateral trade linkages would suggest, as they are amplified by third-country effects (i.e. via linkages with other economies outside the bilateral relationship). For example, Germany benefits not only from the direct spillover from higher Italian demand but also from the increased economic activity of Italy’s other trading partners, which themselves require imports from Germany to grow.

Table 3.

Cross-country effects of (counterfactual) unilateral plans and NGEU

Note: This table displays cross-country GDP effects after 6 years of the counterfactual unilateral investment plans (by row) on the other countries (by column). For example, the cell in row DE and column BE shows that the unilateral German stimulus plan would increase the level of Belgian GDP by 0.06%, while the cell (BE, BE) shows domestic GDP effects in Belgium of the Belgian investment stimulus alone. Two-letter country codes follow EU conventions (https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Country_codes). The last row shows the effects of the synchronised NGEU stimulus. Small differences between the column sums and the NGEU effects relate to model non-linearities. All simulations assume a six-year implementation. Table I.1 in Online Appendix I reports corresponding results for a fast implementation scenario.

The final row shows the total effects of NGEU for each Member State from the simulation, including all NGEU spending jointly. Looking at the effects per country, one sees that the overall GDP effects for small open economies that receive a small direct allocation of funds can be considerably enlarged by the spillovers from other countries. For example, the direct impact for Belgium is small, but spillovers from other countries to Belgium are more than double this effect.

Graph 7 shows that trade is a central mechanism explaining the effects. For each unilateral investment plan, the graph reports the relation of import shares (as a percentage of GDP) and spillover effects (as in Table 3). Overall, there is a strong relationship between import shares and spillovers, and the output gains are most significant for close trading partners. For example, the Belgian investment increases Luxembourgish GDP the most, while the Czech plan mostly affects the Slovakian economy. Across the different panels and as discussed above, the largest plans (e.g. Italy) have the most pronounced spillover effects on other countries. Besides bilateral trade, the broader production network (third-country effects) and the monetary policy setting also determine spillover effects in the model. The following section will analyze these channels in more detail.

Graph 7.

Bilateral trade and spillover effects. Note: This reports the relation of import shares (vertical axis, in percentage of GDP, see also Online Appendix C, Table C.3) and spillover effects. Each panel shows the GDP effects (in %) after 6 years of the counterfactual unilateral investment plans on the other countries (as reported in Table 3). Solid lines show a linear least-squares fit. Note that the x-scales are different across panels. Two-letter country codes follow EU conventions (https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:Country_codes ). All simulations assume a 6-year implementation.

5.6. Spillover channels: the case of Italy

Our main finding is that the spillovers of a coordinated investment stimulus can be sizeable in the EU. We now shed more light on the different transmission channels. Direct trade (export demand) is arguably a major transmission channel for fiscal spillover in the highly integrated European economy, as illustrated in Graph 7. However, there are other potential sources: euro-area members share a common monetary policy and exchange rate. Moreover, countries can benefit from fiscal spillovers via third-country effects even if direct bilateral trade linkages are minor. Given the importance of Italy, the largest recipient country of NGEU funds, the analysis focuses on the (stylized) Italian investment plan.

We contrast the baseline simulation for the Italian investment stimulus presented above with two counterfactuals. First, a “no-bilateral-trade” counterfactual investigates the role of direct trade. We reconsider the Italian public investment but shut down bilateral trade links with Italy one country at a time.Footnote 35 This approach gives 26 simulations, one for each EU trade partner of Italy. While the bilateral trade is turned off, other channels remain, such as common monetary policy and third-country effects. A second counterfactual “no-trade” simulation shuts down (almost) all Italian trade linkages and thereby shuts down third-country effects entirely.Footnote 36 At the same time, the scenario maintains the common monetary policy, which targets euro-area-wide variables. Table 4 summarizes the different assumptions.

Table 4.

Overview of simulation assumptions: the case of Italy

Graph 8.

Spillover effects (in % of GDP) of the Italian investment under different scenarios. Note: This graph displays average GDP effects (in %, 2021–2026) for the baseline and different counterfactual model settings. All results consider the Italian NGEU investment only. All simulations assume a 6-year implementation and high productivity of public investment.

Graph 8 summarizes the results. First, note that sizeable bilateral spillover remains even without direct trade. While perhaps surprising, this finding can be explained by third-country effects, the interest-rate channel, and the exchange-rate channel. Since Italy’s NGEU program increases export demand for all trading partners, positive output spillovers are only partially reduced. Moreover, even if the demand for imported goods is absent, euro-area members benefit from a favorable reaction in real interest rates (see the discussion on the effective lower bound above) and an exchange-rate depreciation. The interaction of these channels generates powerful spillover effects: higher spillovers and output effects throughout the EU amplify the transmission via financial markets. As discussed above, with larger GDP effects in each country, this amplification is strongest for a coordinated expansion.

Our second counterfactual underlines the importance of these interaction effects. When Italy acts as a closed economy in a monetary union, we obtain negative spillovers. Absent trade links, there is no additional demand for foreign goods, but the increase in the real interest rate crowds out demand. Interest-rate differentials then explain the reduced spillovers, in line with Hebous and Zimmermann (Reference Hebous and Zimmermann2013).Footnote 37 As expected, this channel is absent outside the euro area, where the negative effects can be explained by adverse spillover from trading partners in the euro area.

Finally, there are other important factors besides the international linkages exemplified here: the NGEU allocation, the size of the domestic multiplier, the full set of trade linkages (including with third countries), and other country-specific parameters that determine bilateral spillover effects.