## cr18224

## Source details

**Canonical URL:** [cr18224](https://www.imf.org/-/media/files/publications/cr/2018/cr18224.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2018/cr18224.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2018/cr18224.pdf.json)

---

### Overview and headline findings
- Two complementary approaches estimate long-term losses from Brexit for EU27: (i) a multidimensional synthetic index capturing depth and evolution of integration (trade via supply chains, financial linkages, migration) and (ii) a standard multi-country and multi-sector computable general equilibrium (CGE) model.
- Estimated long-run impact on EU27 output: falls by between 0.06 and up to 1.5 percent in the long run (range depends on “soft” vs “hard” Brexit and channels accounted for).
- Substantial heterogeneity across countries; Ireland, Netherlands, and Belgium are among the most affected in the simulations.

### Dimensions of integration: trade
- Total trade in goods and services between the euro area and the U.K. accounts for about 6 percent of euro area GDP (average over the past two decades).
- Euro area trade surplus with the U.K. reached 1 percent of EA GDP in 2016.
- Trade exposure by country: most significant (relative to country size) for Ireland, the Netherlands, Belgium, and Luxembourg.
- Intermediate inputs composition:
  - over 50 percent of goods trade is in intermediate inputs;
  - almost 70 percent of services trade is in intermediate inputs.
- Domestic value-added (DVA) trade exposure to the U.K.: about 3.7 percent of EA GDP, representing about 65 percent of the exposure in gross trade terms.
- In value-added terms, Ireland, Luxembourg, Netherlands exhibit highest exposure to the U.K.

### Supply chain linkages
- Supply chains create indirect links: a significant share of exports involves value added transiting through third countries or re-exported via the U.K.
- Direct value-added exports dominate, but re-exports and transits through third countries are non-negligible for several countries.

### Financial linkages
- Euro area total financial claims and liabilities with the U.K.: about 55 percent of euro area GDP in 2016.
- Key bilateral financial positions (relative to country GDP, avg. 2014–16):
  - Two-way FDI stock between Netherlands and U.K.: about 120 percent of Netherlands’ GDP.
  - Two-way portfolio investments between Ireland and U.K.: slightly below 230 percent of Ireland’s GDP.
  - Two-way bank claims between Luxembourg and U.K.: about 220 percent of Luxembourg’s GDP.
- Net terms: euro area provides financial capital to the U.K. amounting to about 9 percent of euro area GDP.
- Cross-country heterogeneities:
  - Netherlands and Ireland contribute most to the net FDI investment position (about 2.1 percent of euro area GDP in 2016).
  - Relative to their own GDPs, Luxembourg and Ireland are large recipients of cross-border bank lending from the U.K. (more than 170 percent of GDP in the case of Luxembourg and 58 percent of GDP in the case of Ireland).
- Total two-way portfolio investment: about 21 percent of EA GDP (aggregate).

### Migration
- Aggregate migration flows between the euro area and the U.K. are small, with country-level exceptions tied to historical links.
- Euro area historically a net sender of migrants to the U.K. for all skill levels, total balance about 0.1 percent of the euro area population as of 2010.
- Migrant stocks: migrants from Ireland, Cyprus, and Malta living in the U.K. account for roughly 10 percent of these countries’ population.
- U.K. migrants in the euro area: one U.K. migrant for every four to five hundred euro area citizens; larger U.K. migrant populations in Ireland, Luxembourg, and Spain.

### Synthetic integration index — methodology and historical pattern
- Construction:
  - Principal component analysis (PCA) on seven variables; first principal component explains 60 percent of total variance and is positively correlated with the seven variables.
  - Index rescaled to range between 0 (minimal exposure) and 10 (highest exposure).
  - Annual data period: 1993–2013.
- Indicators included:
  - Trade in domestic value added (Ignatenko et al. (2017)).
  - Participation in supply chains (Ignatenko et al. (2017)).
  - Openness in service trade (Ignatenko et al. (2017)).
  - Cross-border banking positions (BIS locational data).
  - Migration share (Brücker et al. (2013)).
- Historical pattern and heterogeneity:
  - Integration index increased by 40 percent over the past 25 years in three phases:
    - First phase: increasing by 20 percent in runup to euro adoption.
    - Second phase: relatively flat after euro adoption.
    - Third phase: increases by another 20 percent in aftermath of global financial crisis.
  - Increase largely driven by Ireland, Belgium, the Netherlands, and Malta; other notable ties: Germany, Finland, Cyprus, Denmark, Sweden.

### Econometric design and long-run elasticities
- Two-step approach:
  1. Regress EU27 output/employment on controls and the integration index.
  2. Assess impact of a decline in integration under different scenarios.
- Panel cointegration techniques account for endogeneity, slow-moving structural variables, and confounding factors.
- Control variables: overall trade openness, domestic investment ratio, inflation rate, total population; country fixed effects included.
- Estimation period: 1993–2013; sample restricted to European countries.
- Long-run semi-elasticities:
  - Long-run semi-elasticity for output: around 0.11.
  - Long-run semi-elasticity for employment: around 0.05.
- Interpretation: a decline in integration (departure from current EU membership arrangements) will negatively affect output and employment in the EU27.

### Calibration of Brexit scenarios via gravity model and econometric impacts
- Gravity model estimates coefficients for arrangement dummies: EU membership (β1), EEA (β2), FTA (β3).
- Derived declines in integration index:
  - (β1 − β2): decline from EU membership to an EEA.
  - (β1 − β3): decline from EU membership to an FTA.
  - (β1): decline from EU membership to the default WTO.
- Calibrated long-run impacts on EU27 (combining index declines with semi-elasticities):
  - EEA scenario (“soft Brexit”): almost zero cost of 0.06 percent for the EU as a whole (output and employment).
  - FTA scenario: output loss of 0.8 percent.
  - WTO (no-deal) scenario: output loss of 1.5 percent.
  - Employment losses across scenarios: range between 0.3 and 0.7 percent.
- Statistical notes and caveats:
  - Estimates statistically significant; Delta method used to assess significance of coefficient differences.
  - Results conditional on statistical power; mask cross-country and cross-sector heterogeneity; do not capture short-run uncertainty or hybrid arrangements.

### CGE model framework, scenarios, and key parameter values
- Model framework and coverage:
  - Armington/monopolistic competition baseline; n countries, each supplies a distinct good; market clearing determines income and trade patterns.
  - Baseline covers 34 countries and 31 sectors; tradable intermediate inputs included to capture supply chain linkages.
  - Three versions: perfect competition, monopolistic competition (primary), firm heterogeneity (Melitz) for smaller sample (10 countries, 16 sectors).
- Data inputs:
  - Trade linkages: WIOD year 2011; aggregated to 34 countries and 31 sectors.
  - MFN tariff data: UK MFN tariffs by sector (Dhingra et al. (2016)).
  - Non-tariff barriers: EU Exit Analysis Cross Whitehall Briefing paper; Berden et al. (2009, 2013).
- Trade elasticity choices:
  - Agriculture and manufacturing: Caliendo and Parro (2015).
  - Service sectors: aggregate trade elasticity of 5 (Costinot and Rodriguez-Clare (2013)).
  - Exceptions: coke, refined petroleum and nuclear fuel sector elasticity close to 0; transport equipment elasticity calibrated to Egger and Kaynak (2017).
- Selected UK MFN tariff (percent, table excerpts):
  - Agriculture, Hunting, forestry and fishing: Imports 5.9 Exports 5.6
  - Mining and quarrying: Imports 0 Exports 0
  - Food, beverages and tobacco: Imports 7.26 Exports 4.96
  - Textiles and textile products, Leather: Imports 9.5 Exports 9.7
  - Transport equipment: Imports 8.09 Exports 7.22
  - Weighted average (by EU trade): Imports 4.43 Exports 3.29

### CGE scenarios and calibration details
- Implementation: increases in goods tariffs and non-tariff barriers for goods and services trade.
- FTA scenario:
  - U.K. leaves single market and customs union but agrees a broad free trade agreement.
  - Tariffs on goods trade remain at zero; non-tariff costs increase moderately across sectors.
  - Financial sector calibration: net exports of financial services from the U.K. to the EU fall by about 40 percent.
  - Transport equipment: assumed higher increase of non-tariff trade costs; robustness checks do not change results significantly.
- Hard Brexit (WTO) scenario:
  - U.K. and EU trade on WTO terms; MFN tariffs applied.
  - Non-tariff trade costs increase by twice as much as in the FTA scenario.
- Transition assumption: smooth transition to new arrangements in both scenarios.

### CGE model quantitative impacts and country results
- Baseline monopolistic competition results — long-term EU output losses:
  - FTA scenario: EU output losses of 0.2 percent.
  - Hard Brexit (WTO) scenario: EU output losses of 0.5 percent.
- Country-specific long-run real income falls:
  - Ireland: about 2.5 to 4 percent (similar to estimated impact on the U.K.).
  - Netherlands: about 0.7 percent in FTA scenario and about 1 percent in WTO scenario.
  - Belgium: about 0.5 percent in FTA scenario and about 1 percent in WTO scenario.
- Summary range across approaches:
  - Pessimistic scenario (staff estimates): range of output loss between of 0.5 and 1.5 percent over the long run.
  - Econometric model tends to point to larger impacts than CGE estimates because it includes broader integration channels beyond trade.

### Distributional and sectoral considerations
- Losses uneven across countries and sectors:
  - Most integrated countries (Ireland, Luxembourg, Netherlands, Belgium, Malta, Cyprus) likely to suffer disproportionately.
  - Sector concentration: 70 percent of aggregate impact falls in five sectors: automotive; agriculture, food & drink; chemicals & plastics; consumer goods; and industrials (estimated 75 percent of the impact in those sectors).
  - Prior findings: “motor vehicles” and “machinery & equipment” potentially most affected in Germany in value added terms.
- Supply chain positioning and substitutability of financial centers (London vs euro area capitals) matter for impact size.

### Robustness, limitations, and validation
- CGE model limitations: linear cost function, Dixit-Stiglitz preferences, sensitivity to trade elasticities, omitted relocation of U.K. subsidiaries of multinationals.
- Validation: CGE models capture tariff-driven changes reasonably well in historical contexts (e.g., Caliendo and Parro (2015)); staff CGE estimates broadly similar to literature (Dhingra et al., Aichele and Felbermayr, OECD, Roja-Romagosa, Booth et al.).
- Econometric and CGE results broadly consistent with studies augmenting CGE with supply chain links (Connell et al., 2017).

### Youth unemployment and labor market findings (euro area recovery context)
- Trajectory and levels:
  - Youth unemployment rate declined by more than 5 percentage points to under 19 percent by 2017, from a peak of 24 percent in 2013.
  - Adult unemployment fell by close to 3 percentage points to 8 percent in 2017.
- Labor force and employment dynamics (age 15–24):
  - Active young population started to decline with onset of global financial crisis in 2008 and stabilized in 2016.
  - Cumulative reduction in young labor force: almost 3 million during 2008–17.
  - More than 3 million jobs lost for the young over the same period.
  - Since 2013, number of unemployed young declined by 0.9 million, but only 0.3 million new jobs created.
- Demographics, migration, and education:
  - Net migration inflows for age 15–24 fell to about 0.2 million annually during 2008–16 (annual average around 0.3 million during 2006–07).
  - Proportion of young people in education rose almost 5 percentage points since 2008, to 57 percent in 2017.
- NEET and job quality:
  - NEET for 15–24 back to pre-crisis level by 2017; group of eight high-unemployment countries had NEET rate 14.1 percent vs 6.6 percent for rest of euro area in 2017.
  - Over 2013–17, total employment grew by 5.4 percent, compared with 3.1 percent for the young.
  - Job creation concentrated in skill-intensive sectors accounted for close to 60 percent of total employment increase.
  - New jobs for the young were almost exclusively part time: part-time jobs grew by about 10 percent for the young over 2013–17; no increase in full-time jobs for the young.
  - Job growth highest for the highly educated: close to 13 percent over 2013–17; almost 19 percent for young workers with high education attainment.
- “3 Million Missing Young”:
  - Over 2008–17, over 3 million young workers lost jobs and active young population reduced by close to 3 million (17 percent of young labor force in 2007).
  - Between 2008–13 youth employment reduced by over 3.3 million and youth labor force reduced by about 2.3 million.

### Empirical analysis of youth employment determinants — key coefficients (EU multivariate estimates)
- Tax wedge effects:
  - Age 15–24: coefficients reported as -0.38*** (SE 0.07), -0.37*** (0.10), 0.70*** (0.13), 0.67*** (0.16) across specifications.
  - Age 25–54: coefficients reported as -0.19* (0.10), -0.24* (0.12), 0.32*** (0.08), 0.38*** (0.08).
  - Interpretation: high labor tax wedge appears particularly harmful for the young; estimated effect for the young is more than double effect on prime-age workers.
- ALMP spending (active labor market policy):
  - Age 15–24: 0.53*** (0.08), 0.43*** (0.06), -0.95*** (0.10), -0.92*** (0.13) across specifications.
  - Age 25–54: 0.35** (0.16), 0.26 (0.16), -0.34*** (0.05), -0.32*** (0.06).
  - Interpretation: ALMPs focusing on training and education boost employment, particularly for the young.
- Net replacement rate:
  - Age 15–24: -0.07** (0.03), -0.07** (0.03), -0.06 (0.05), -0.05 (0.05).
  - Age 25–54: -0.03 (0.03), -0.02 (0.04), -0.01 (0.02), -0.01 (0.02).
- Coordination of wage setting:
  - Coefficient reported as 0.64* (0.34) for age 15–24 and 0.50 (0.31) for age 25–54; alternate specifications include -0.61 (0.51) and -0.58* (0.27).
- Model fit and sample:
  - Number of observations: 354 and 300 across specifications.
  - R-squared ranges: 0.605 to 0.789.

### Policy recommendations (labor market, fiscal, structural)
- Labor market reforms:
  - Tackle labor market duality and ensure efficient collective bargaining.
  - Address skill mismatches and retraining through well-designed apprenticeship systems and ALMPs.
  - Provide adequate social protection that adapts to a changing job market.
- Fiscal policy:
  - Reduce labor tax wedge.
  - Ringfence and increase the efficiency of education spending.
  - Target education spending to high-labor-demand and high-productivity areas.
- Product market reforms:
  - Improve business environment by cutting red tape and opening up regulated professions.
  - Promote further integration within the EU to benefit from economies of scale from the single market.
  - Deepen financial markets and improve personal insolvency laws to promote entrepreneurship and innovation.

*Source: IMF staff estimates and analysis (content unit cr18224).*

### 1. Trade in Goods, Services, and Financial Services ________________________________________ 6

### 1. Trade in Goods, Services, and Financial Services

### Overview and headline findings
- The paper uses two approaches to estimate long-term losses from Brexit for EU27: (i) a multidimensional synthetic index capturing depth and evolution of integration (trade via supply chains, financial linkages, migration) and (ii) a standard multi-country and multi-sector computable general equilibrium (CGE) model.
- Estimated long-run impact on EU27 output: falls by between 0.06 and up to 1.5 percent in the long run. The range depends on whether a “soft” or “hard” Brexit is assumed, and whether trade or other transmission channels are accounted for.
- Substantial heterogeneity across countries; Ireland, Netherlands, and Belgium are among the most affected in the simulations.

### Dimensions of integration: trade
- Total trade in goods and services between the euro area and the U.K. accounts for about 6 percent of euro area GDP (average over the past two decades).
- The euro area runs a modest trade surplus with the U.K.; the euro area’s trade surplus with the U.K. increased steadily, reaching 1 percent of EA GDP in 2016.
- Trade with the U.K. is most significant (relative to country size) for Ireland, the Netherlands, Belgium, and Luxembourg.
- Most trade is in intermediate inputs:
  - over 50 percent of goods trade is in intermediate inputs;
  - almost 70 percent of services trade is in intermediate inputs.
- Domestic value-added (DVA) trade exposure to the U.K. has been about 3.7 percent of EA GDP, representing about 65 percent of the exposure in gross trade terms.
- In value-added terms, smaller but open economies (Ireland, Luxembourg, Netherlands) exhibit the highest exposure to the U.K., though exposure is smaller than suggested by gross trade statistics.

### Supply chain linkages
- Supply chains create indirect links: a significant share of exports involves value added transiting through third countries or re-exported via the U.K.
- Direct value-added exports dominate, but re-exports and transits through third countries are non-negligible for several countries.

### Financial linkages
- Euro area total financial claims and liabilities with the U.K. amounted to about 55 percent of euro area GDP in 2016.
- Key country-level bilateral financial positions (relative to country GDP, avg. 2014–16 and examples cited):
  - Two-way FDI stock between Netherlands and U.K. is about 120 percent of Netherland’s GDP.
  - Two-way portfolio investments between Ireland and U.K. is slightly below 230 percent of Ireland’s GDP.
  - Two-way bank claims between Luxembourg and U.K. is about 220 percent of Luxembourg’s GDP.
- In net terms, the euro area provides financial capital to the U.K. amounting to about 9 percent of euro area GDP.
- Cross-country heterogeneities in net positions:
  - Netherlands and Ireland contribute most to the net FDI investment position (about 2.1 percent of euro area GDP in 2016).
  - Ireland and Malta have large net portfolio investment positions with the U.K., whereas most other countries are net recipients.
  - Relative to their own GDPs, Luxembourg and Ireland are large recipients of cross-border bank lending from the U.K. (more than 170 percent of GDP in the case of Luxembourg and 58 percent of GDP in the case of Ireland).
- Total two-way portfolio investment amounts to about 21 percent of EA GDP (aggregate).

### Migration
- Migration flows between the euro area and the U.K. are small in aggregate, with important country-level exceptions tied to historical links.
- The euro area has traditionally been a net sender of migrants to the U.K. for all skill levels, with a total balance of about 0.1 percent of the euro area population as of 2010.
- The number of migrants from Ireland, Cyprus, and Malta living in the U.K. is considerable, accounting for roughly 10 percent of these countries’ population.
- Regarding migration from the U.K. to the euro area: there is one U.K. migrant living in the euro area for every four to five hundred euro area citizens; the U.K. migrant population is larger in Ireland, Luxembourg, and Spain.

### Stylized implications for Brexit
- The strength and multidimensional nature of euro area–U.K. integration implies there would be no Brexit winners; both sides would incur losses.
- Higher barriers to trade, capital flows, and people movements following Brexit could disrupt links and reduce trade, investment, and labor mobility.
- Empirical studies reviewed in the paper suggest average long-term impacts on EU27 output in various studies:
  - A review finds an average long-term impact of Brexit on EU27 output between -0.2 and -0.5 percent by 2030.
  - Connell et al. (2017) find an impact of Brexit on the EU in the order of -0.4 percent (for the ‘soft Brexit’ scenario) and -1.4 percent (for the ‘hard Brexit’ scenario).
  - Oliver Wyman and Clifford Chance (2018) find annual “red tape” costs of about £31 billion (0.3 percent of EU27 GDP) for EU27 exporters; a future customs arrangement equivalent to the Customs Union reduces the EU27 impact to around £14 billion (0.13 percent of EU27 GDP).
  - Sector concentration: the report finds that 70 percent of the aggregate impact falls in five sectors: automotive; agriculture, food & drink; chemicals & plastics; consumer goods; and industrials (estimated 75 percent of the impact in those sectors).
  - Chen et al. (2018) find regions in Ireland, Malta, Netherlands, Belgium, and Germany most likely to be affected.
- The paper’s own CGE and index-based estimates: EU27 output falls by between 0.06 and up to 1.5 percent in the long run, with heterogeneity across countries (Ireland exhibiting the highest exposure; also Luxembourg, Netherlands, Belgium, Malta, Cyprus; Germany can be affected via supply chains).

### Synthetic index of exposure to the U.K.
- The paper constructs a single country-specific, time-varying synthetic index aggregating subcomponents of integration (trade via supply chains, financial linkages, migration).
- Purpose and advantages:
  - Captures the multiple correlated dimensions of the EU27–U.K. economic relationship in a single measure.
  - Solves multicollinearity concerns that would arise if all components were used separately in regressions.
  - Used in regressions to assess effects on euro area output and employment from integration with the U.K.

### Distributional and sectoral considerations
- Losses are unevenly distributed across countries and sectors:
  - Countries more integrated with the U.K. (Ireland, Luxembourg, Netherlands, Belgium, Malta, Cyprus) will likely suffer disproportionately from Brexit.
  - Industrial sectors such as “motor vehicles” and “machinery & equipment” identified in prior work (Connell et al. (2017)) as potentially most affected in Germany in terms of value added.
  - Supply chain positioning and the degree of substitutability between London and euro area capitals as financial centers matter for impact size.

*Source: IMF — Long-Term Impact of Brexit on the EU (June 29, 2018), Chapter 1: "Trade in Goods, Services, and Financial Services".*

### 8.      To build the integration index, we use a principal component analysis. Principal

### 8.      To build the integration index, we use a principal component analysis. Principal

### Methodology: construction of the integration index
- Principal component analysis (PCA) is used to resolve dimensionality and multicollinearity from multiple indicators of bilateral economic integration with the U.K.
- The exposure index is rescaled to range between 0 (minimal exposure) and 10 (highest exposure).
- Annual data covering the period 1993–2013 are used.
- The first principal component explains 60 percent of the total variance and is positively correlated with the seven variables used to build the exposure index.
- Integration indicators included (data sources indicated where provided):
  - Trade in domestic value added: sum of bilateral exports and imports of domestic value added normalized by the country’s GDP. Data are based on Ignatenko et al. (2017).
  - Participation in supply chains: sum of “backward” and “forward” trade linkages between each country and the U.K., normalized by the country’s GDP. Data are from Ignatenko et al. (2017).
  - Openness in service trade: sum of each country’s exports of services to, and imports of services from, the U.K., normalized by GDP. Data are from Ignatenko et al. (2017).
  - Cross-border banking positions: ratio of claims by international banks in the U.K. on each receiving country from BIS locational data, normalized by GDP.
  - Migration: share of each country’s migrants residing in the U.K., normalized by the country’s total number of migrants residing in 20 OECD countries. Data are from Brücker et al. (2013). Migrants are defined as foreign-born individuals aged 25 years and older, living in each of the 20 considered OECD destination countries.
- Bilateral FDI and portfolio statistics were not retained due to data availability; cross-border banking flows are used to capture financial-account bilateral integration.

### Integration index: historical pattern and cross-country heterogeneity
- The integration index indicates euro area–U.K. integration strengthened over the years, with the intensity of integration increasing by 40 percent over the past 25 years, split in three phases:
  - First phase: increasing by 20 percent in the runup to euro adoption.
  - Second phase: after euro adoption, the index stayed relatively flat.
  - Third phase: in the aftermath of the global financial crisis, integration increases by another 20 percent.
- Increased integration is driven largely by a handful of countries: Ireland, Belgium, the Netherlands, and Malta.
- Other countries with considerable ties to the U.K.: Germany, Finland, Cyprus, and non-euro area countries such as Denmark and Sweden.

### Econometric design for long-run effects on EU27 output and employment
- Objective: assess long-term effect on EU27 output and employment of Brexit modeled as a partial reversal of EU27 integration with the U.K.
- Two-step approach:
  1. Estimate relationship between EU27 output/employment dynamics and their integration with the U.K., regressing output/employment on controls and the integration index.
  2. Assess impact on output/employment of a decline in integration under different future relationship scenarios between the U.K. and EU27.
- Panel cointegration techniques are used to estimate long-run effects, accounting for:
  - Potential endogeneity of the integration index.
  - Slow-moving structural variables implying long-horizon effects.
  - Confounding with EU Single Market membership or overall trade openness.
- Control variables: overall trade openness (total exports plus imports over GDP), domestic investment ratio, inflation rate, total population.
- Country fixed effects are included to capture time-invariant or slow-moving influences.
- Annual macro variables drawn from the IMF World Economic Outlook and World Development Indicators databases.
- Model estimated for the period 1993–2013; sample restricted to European countries.

### Econometric findings: long-run elasticities and implications
- Positive long-run relationship between bilateral integration with the U.K. (synthetic index) and EU27 outcomes:
  - Long-run semi-elasticity for output: around 0.11.
  - Long-run semi-elasticity for employment: around 0.05.
- Interpretation: a decline in the level of integration, via departure from current EU membership arrangements, will negatively affect output and employment in the EU27.

### Calibration of Brexit scenarios via a gravity model
- A gravity model is used to estimate the effects of different economic arrangements on the integration index, controlling for bilateral distance, common border, common language, regional dummies, population size, GDP level, and year fixed effects.
- Three arrangement dummies are estimated:
  - EU membership (β1): effect of EU membership on the integration index.
  - European Economic Area (EEA) arrangement (β2): impact of EEA membership (e.g., Norway, Iceland).
  - Standard free trade agreement (FTA) (β3): effect of an FTA.
- From estimated coefficients, the study derives reductions in the integration index consistent with scenarios:
  - (β1 − β2): decline going from EU membership to an EEA.
  - (β1 − β3): decline going from EU membership to an FTA.
  - (β1): decline from EU membership to the default WTO.

### Calibrated long-run impacts of Brexit scenarios on EU27
- Combining estimated declines in the integration index with the long-run semi-elasticities yields long-run impacts:
  - EEA scenario (preserve single market access, sacrifice customs union; “soft Brexit”): almost zero cost of 0.06 percent for the EU as a whole, for both output and employment.
  - FTA scenario: output loss of 0.8 percent.
  - WTO (no-deal) scenario: output loss of 1.5 percent.
  - Employment losses across scenarios range between 0.3 and 0.7 percent.
- Remarks:
  - These estimates are on average higher than previous studies using standard CGE trade models, but broadly similar to studies augmenting CGE models with supply chain links (e.g., Connell et al., 2017).
  - The econometric approach incorporates additional channels of integration beyond trade.

### Statistical precision, caveats, and extensions
- The estimated effects on integration and resulting impacts are statistically significant with relatively good precision; the Delta method is used to assess statistical significance of coefficient differences.
- Results should be interpreted with caution:
  - They are conditional on the statistical power of tests and represent average effects for the EU.
  - Econometric estimations are subject to statistical uncertainty.
  - Results mask cross-country and cross-sector heterogeneity reflecting different exposures to the U.K.
  - Economic uncertainty surrounding post-Brexit arrangements may have short-run effects not captured here.
  - Scenarios considered are polar and rigid; hybrid arrangements between the EU and U.K. are not modeled due to limited data variation for interactions.
- Next methodological step (introduced but not detailed in this excerpt): use of a multi-country, multi-sector CGE model to quantify country-by-country and sector-by-sector effects from higher trade barriers, capturing heterogeneous sectoral exposures to the U.K.

*Source: IMF staff estimates and analysis (content unit cr18224).*

### 19.      The core of the model is to infer changes in real income associated with changes in

### cr18224 - 19.      The core of the model is to infer changes in real income associated with changes in

### Model framework and core intuition
- Armington model (simple CGE): n countries, each supplies a distinct good; representative household maximizes utility subject to budget constraint; trade flows determined by preferences, income, cost of trade (tariffs), and foreign prices.
- Market clearing: demand for any good must equal supply; changes in trade costs imply solving for new pattern of income changes consistent with new bilateral trade costs.
- Key mechanism: an increase in trade cost reduces export revenues for affected exporters, lowers income, triggers knock-on effects to other countries (even if their trade costs do not change), and requires imports to fall to maintain external balance; households face lower income and fewer varieties consumed.
- Extensions: insights from Armington carry into more complex frameworks with monopolistic competition and firm heterogeneity.

### CGE model specification (baseline and variants)
- Baseline covers 34 countries and 31 sectors, includes tradable intermediate inputs to capture global supply chain linkages, and assumes monopolistic competition among firms for the focal results.
- Three model versions:
  - Perfect competition: multiple countries and sectors (34 countries plus rest of world, 31 sectors); provides lower bound to welfare effects of changes in trade costs.
  - Monopolistic competition: Krugman (1980) style product differentiation; used for primary discussion.
  - Firm heterogeneity: Melitz (2003) style, but implemented for a smaller set (10 countries and 16 sectors) due to computational burden.
- Outcome measured: changes in real income (consumption and welfare) comparing scenario where U.K. remains EU member versus U.K. does not.

### Data inputs and parameter choices
- Trade linkage data: World Input-Output Database (WIOD), year 2011; aggregated into 34 countries, rest of world, and 31 sectors.
- MFN tariff data: applied most favored nation (MFN) tariff by the EU from Dhingra et al. (2016) for 31 sectors.
- Non-tariff trade barriers: primary source—EU Exit Analysis Cross Whitehall Briefing paper; supplemented with Berden et al. (2009, 2013) estimates (tariff-equivalent of non-tariff barriers between U.S.A. and EU).
- Trade elasticities:
  - Agriculture and manufacturing sectors: Caliendo and Parro (2015).
  - Service sectors: held equal to aggregate trade elasticity of 5 following Costinot and Rodriguez-Clare (2013).
  - Exceptions: trade elasticity on coke, refined petroleum and nuclear fuel sector set to close to 0; transport equipment elasticity calibrated in line with Egger and Kaynak (2017).
- Table 1 (UK MFN Tariff With Non-EU Countries) — selected entries (percent):
  - Agriculture, Hunting, forestry and fishing: Imports 5.9 Exports 5.6
  - Mining and quarrying: Imports 0 Exports 0
  - Food, beverages and tobacco: Imports 7.26 Exports 4.96
  - Textiles and textile products, Leather: Imports 9.5 Exports 9.7
  - Transport equipment: Imports 8.09 Exports 7.22
  - Weighted average (by EU trade): Imports 4.43 Exports 3.29

### Alternative post-Brexit scenarios modeled
- Implementation: modeled as increases in goods tariffs and non-tariff barriers for goods and services trade.
- 'FTA' scenario:
  - U.K. leaves single market and customs union but agrees a broad free trade agreement with EU.
  - Tariffs on goods trade remain at zero.
  - Non-tariff costs increase moderately across sectors.
  - Financial sector: non-tariff trade cost calibrated so net exports of financial services from the U.K. to the EU fall by about 40 percent.
  - Transportation equipment sector: assumed higher increase of non-tariff trade costs (to reflect complicated supply chains); robustness checks with lower tariffs on transportation equipment do not change results significantly.
- 'Hard Brexit' (WTO) scenario:
  - U.K. trades with EU on WTO terms; U.K. applies MFN tariffs on goods imported from EU and EU applies MFN tariffs on goods originating from U.K.
  - Non-tariff trade costs increase by twice as much as in the FTA scenario for all sectors.
- Transition assumption: in both scenarios, assume the U.K. and EU transition smoothly to the new trading arrangement.

### Results — quantitative impacts
- Baseline model (monopolistic competition) long-term EU output losses:
  - FTA scenario: EU output losses of 0.2 percent.
  - Hard Brexit (WTO) scenario: EU output losses of 0.5 percent.
- Country-specific estimates (long run):
  - Ireland: real income fall of about 2.5 to 4 percent (similar to estimated impact on the U.K.).
  - Netherlands: real income fall of about 0.7 percent in FTA scenario and about 1 percent in WTO scenario.
  - Belgium: real income fall of about 0.5 percent in FTA scenario and about 1 percent in WTO scenario.
- Summary range across empirical approaches (paragraph 25):
  - Pessimistic scenario: staff estimates suggest range of output loss between of 0.5 and 1.5 percent over the long run.
  - Econometric model tends to point to larger impacts than CGE model-based estimates because it considers broader channels beyond trade.

### Robustness, limitations, and caveats
- Model assumptions limiting representation of reality:
  - Linear cost function and Dixit-Stiglitz preferences.
  - Uncertainty and sensitivity to assumed trade elasticities; estimation of trade elasticities is econometrically challenging with significant variation across studies.
- Omitted channels:
  - Potential relocation of U.K. subsidiaries of multinational firms not considered.
- Validation and literature comparison:
  - Caliendo and Parro (2015) show CGE model performs reasonably well capturing tariff-driven changes (example: NAFTA 1993–2005).
  - CGE models remain cornerstone for trade policy evaluation (Baldwin and Venables, 1995; Piermartini and Teh, 2005).
  - Staff CGE model estimates broadly similar to literature focusing on trade effects of Brexit (Dhingra et al., 2016; Aichele and Felbermayr, 2015; OECD, 2016; Roja-Romagosa, 2016; Booth et al., 2016).
  - Connell et al. (2017) using a deeper CGE model with complex supply chain linkages produce similar results for EU27 to those from the econometric model.

### Key figures referenced
- Model coverage: 34 countries and 31 sectors (baseline).
- Service sector aggregate trade elasticity: 5.
- Financial services net exports from U.K. to EU fall calibration in FTA: about 40 percent.
- Weighted average UK MFN tariff (by EU trade): Imports 4.43 Exports 3.29 (percent).
- Long-run EU output losses: FTA 0.2 percent; WTO 0.5 percent.
- Country impacts cited: Ireland 2.5 to 4 percent; Netherlands 0.7 percent (FTA) and about 1 percent (WTO); Belgium 0.5 percent (FTA) and about 1 percent (WTO).
- Pessimistic scenario output loss range (staff econometric model): between of 0.5 and 1.5 percent over the long run.

*Source: https://www.imf.org/-/media/files/publications/cr/2018/cr18224.pdf*

### 27.      This paper has examined the consequences of Brexit on the EU27 under various post-

### cr18224 - 27.      This paper has examined the consequences of Brexit on the EU27 under various post-

### Brexit analysis: approach and headline findings
- The paper examined the consequences of Brexit on the EU27 under various post-Brexit scenarios and using two different, complementary, approaches.
- Results are described as broadly in line with recent findings in the literature.
- Two principal findings:
  - First, Brexit would have negative effects on the EU27 given the depth and the complexity of the EU-U.K. integration.
  - Second, there is significant cross-country heterogeneity in exposure to Brexit-related shocks.

### Quantitative estimates of long-term impacts (EU27)
- Estimated long-term output losses (in percent) for the EU27:
  - Output: fall at most by up to 1.5 percent in the long run in the event of a ‘hard’ Brexit scenario.
  - Employment: fall at most by up to 0.7 percent in the long run in the event of a ‘hard’ Brexit scenario.
- A ‘soft’ Brexit outcome would lead to much lower losses.
- The estimated long-term output and employment losses (in percent) for the EU27 in this study are on average lower than the corresponding losses for the U.K. estimated in the literature (references cited).

### Cross-country heterogeneity and most exposed economies
- Very open economies such as Ireland, the Netherlands, and Belgium are among the most exposed to Brexit-related adverse shocks.
- Ireland is identified as the only EU27 country exhibiting Brexit-related output losses of similar magnitude to those estimated for the U.K. in the literature.

### Appendix — Principal Component Analysis Results (selected quantitative outputs)
- Table A1. Eigen Value and Cumulative Relative Frequencies
  - Principal component 1: Eigen Values 2.99; Proportion 0.6; Cumulative relative frequencies 0.6
  - Principal component 2: Eigen Values 0.98; Proportion 0.2; Cumulative relative frequencies 0.8
  - Principal component 3: Eigen Values 0.88; Proportion 0.17; Cumulative relative frequencies 0.97
  - Principal component 4: Eigen Values 0.10; Proportion 0.02; Cumulative relative frequencies 1.0
  - Principal component 5: Eigen Values 0.04; Proportion 0.0; Cumulative relative frequencies 1.0
- Table A2. Eigen Vectors (P1)
  - Trade in value added-to-GDP: 0.57
  - Participation in supply chains: 0.54
  - Service trade openness: 0.56
  - Cross-border bank claims: 0.1
  - Migration share: 0.27

### Youth unemployment during the euro area economic recovery — key findings
- Overall trajectory and levels:
  - The youth unemployment rate for the euro area declined by more than 5 percentage points to under 19 percent by 2017, from its peak of 24 percent in 2013.
  - Adult unemployment fell by close to 3 percentage points, to 8 percent in 2017.
- Labor force and employment dynamics (young, age 15–24):
  - The active young population started to decline with the onset of the global financial crisis in 2008 and only stabilized in 2016.
  - The cumulative reduction in the young labor force amounted to almost 3 million during 2008–17.
  - More than 3 million jobs were lost for the young over the same period.
  - Since 2013, the number of unemployed young declined by 0.9 million, but only 0.3 million new jobs were created, implying a loss of more than 0.6 million unemployed people in the young labor force (i.e., more than two thirds of the decline in the number of unemployed was due to a shrinking labor force).
- Population, migration, and education effects:
  - The primary reason for the decline in the young labor force was the smaller size of new youth cohorts, exacerbated by lower immigration following the crisis.
  - Net migration inflows for age 15–24 fell to about 0.2 million annually during 2008–16, compared to an annual average of around 0.3 million during 2006–07.
  - The proportion of young people in education rose almost 5 percentage points since 2008, to 57 percent in 2017.
- NEET (not in employment, education or training) trends:
  - The share of 20–34 years old that are NEET rose after 2008 and then started to decline gradually; the NEET rate was back to the pre-crisis level for the 15–24 age group by 2017.
- Job creation composition and quality:
  - Over 2013–17, total employment grew by 5.4 percent, compared with 3.1 percent employment growth for the young.
  - Job creation concentrated in skill-intensive sectors (ICT, finance, energy, professional services, health) accounted for close to 60 percent of total employment increase.
  - Sectors traditionally concentrating youth employment—construction, accommodation, manufacturing, retail trade—accounted for less than 45 percent of total employment increase; the young experienced job loss in construction and retail trade.
  - New jobs for the young were almost exclusively part time: part-time jobs grew by about 10 percent for the young over 2013–17, while there was no increase in full-time jobs for the young.
  - Employment growth by education level:
    - Job growth was highest for the highly educated, at close to 13 percent over 2013–17.
    - The increase was almost 19 percent for young workers with high education attainment.
    - Prime-age workers accounted for over 73 percent of employment growth among the highly educated.
    - Older workers (55+ age group) gained a large share of jobs for low- and medium-education levels.
- Labor market competition effects:
  - Extended working lives and increased labor force participation among older workers, and pension reforms raising retirement ages in several countries, may have increased short-run competition for vacancies and adversely affected youth hiring during the downturn.

*International Monetary Fund — EURO AREA POLICIES (content unit cr18224 - 27)*

### 11.      An empirical analysis shows that several labor market features play a role in youth

### 11.      An empirical analysis shows that several labor market features play a role in youth employment.

### Model and data
- Panel multivariate model (similar to Banerji et al. (2014)) estimating impact on youth (age 15–24) versus prime-age (age 25–54) employment and unemployment rates.
- Model specification (as presented):
  - ݁
    ௜,௧
    ߚ = ଵ,௜
    ߚ	൅ ଶ,௜
    ݕ൫	
    ௜,௧
    ݕെ
    ௜,௧
    ∗
    ൯൅		Σ
    ௝
    ߛ
    ௝
    ݔ
    ௜,௝,௧
    ߳	൅
    ௜,௧
- Variables:
  - ݁
    ௜,௧
    : employment rate or unemployment rate for country ݅ at time ݐ for age group ௧ (youth or prime-age).
  - ݕሺ
    ௜,௧
    ݕെ
    ௜,௧
    ∗
    ሻ : country-specific output gap.
  - ݔ
    ௜,௝,௧
    : labor market feature ݆ in country ݅ at year ݐ.
- Estimation period: 2000–16.
- Panel controls include country-specific output gap coefficient and country fixed effects; Driscoll-Kraay standard errors reported.
- Note: Employment rate is calculated as the ratio of employed over total population for respective age groups.

### Key empirical findings (Table 1: EU multivariate estimates)
- Tax wedge effects (employment and unemployment regressions):
  - Age 15–24: Tax wedge coefficients reported as -0.38*** (SE 0.07), -0.37*** (0.10), 0.70*** (0.13), 0.67*** (0.16) across specifications.
  - Age 25–54: Tax wedge coefficients reported as -0.19* (0.10), -0.24* (0.12), 0.32*** (0.08), 0.38*** (0.08) across specifications.
  - Interpretation: A high labor tax wedge appears particularly harmful for the young; estimated effect for the young is more than double the effect on prime-age workers when looking at both employment and unemployment rates.
- ALMP spending (active labor market policy):
  - Age 15–24: 0.53*** (0.08), 0.43*** (0.06), -0.95*** (0.10), -0.92*** (0.13) across specifications.
  - Age 25–54: 0.35** (0.16), 0.26 (0.16), -0.34*** (0.05), -0.32*** (0.06).
  - Interpretation: ALMPs that focus on training and education boost both young and prime-age employment, and appear particularly beneficial for the young.
- Net replacement rate:
  - Age 15–24: -0.07** (0.03), -0.07** (0.03), -0.06 (0.05), -0.05 (0.05).
  - Age 25–54: -0.03 (0.03), -0.02 (0.04), -0.01 (0.02), -0.01 (0.02).
  - Interpretation: Higher net replacement rate can provide disincentives to work; associated decrease in employment rate is significant for the young, while unemployment regressions are not significant.
- Coordination of wage setting:
  - Coefficient reported as 0.64* (0.34) for age 15–24 and 0.50 (0.31) for age 25–54; other reported values include -0.61 (0.51) and -0.58* (0.27) in alternate specifications.
  - Robustness: Including coordination of wage setting index as an additional control does not seem to affect main findings (Table 1, columns 3–4, 7–8).
- Model fit and sample:
  - Number of observations: 354 (columns with full sample) and 300 (alternate specifications).
  - R-squared ranges reported: 0.605 to 0.789 across specifications.

### Macroeconomic implications
- Temporary contracts and vulnerability:
  - A larger proportion of temporary contracts makes the young more vulnerable to downturns than adult workers.
  - Share of temporary workers in the euro area is high compared to other advanced economies and continued to rise after the crisis, particularly in the YU8 where more than half of the young workers are on temporary contracts.
  - Temporary contracts usually offer less security and may make the young more vulnerable in downturns.
  - Possible negative productivity effects: temporary contracts may reduce employees’ effort or decrease provision of on-the-job training, damaging career prospects.
- Unemployment benefits coverage:
  - Unemployment benefits offer only limited support for the young.
  - Eligibility criteria (duration of employment and duration of benefits) mean many young people are not covered.
  - The share of unemployed young receiving unemployment benefits has declined since 2007 and is substantially lower than the corresponding share for older unemployed.
- Poverty and long-term effects:
  - The number of young people at risk-of-poverty continues to rise in the euro area.
  - Consequences of high youth unemployment, increased discouraged workers, and precarious employment include higher at-risk-of-poverty rates for the young and potential hysteresis effects on skills and incomes, harming human capital and long-term growth potential.

### Quantitative snapshots and demographics
- Youth in 2017:
  - A little over 2.6 million unemployed young in the euro area in 2017; roughly 2 million or more than 77 percent reside in eight euro area countries that account for 61 percent of the euro area young population.
- NEET and disparities:
  - NEET (not in employment, education, or training) in percent of total population in the euro area fell by close to 2 percentage points from the crisis peak of 13 percent.
  - The group of eight high-unemployment euro area countries had NEET rate at 14.1 percent as a group, compared with 6.6 percent for the rest of euro area countries in 2017.
- "3 Million Missing Young" (Box 2):
  - Over the period 2008–17, over 3 million young workers lost their jobs, accompanied by a close to 3 million reduction in the active young population, or 17 percent of the young labor force in 2007.
  - Between 2008–13 youth employment was reduced by over 3.3 million and the youth labor force was reduced by about 2.3 million.
  - Young labor force shrank by 17 percent, compared with a decline of young population by 8 percent over 2008–17.
  - Counterfactual analysis: adverse demographic trend is main driver of shrinking young labor force, but many young either started investing in education or became discouraged.

### Policy recommendations (Deeper structural reforms to boost youth labor market prospects)
- Labor market reforms:
  - Tackle labor market duality and ensure efficient collective bargaining.
  - Address skill mismatches and retraining through well-designed apprenticeship systems and ALMPs.
  - Provide an adequate social protection system that adapts to a changing job market.
- Fiscal policy:
  - Reduce labor tax wedge.
  - Ringfence and increase the efficiency of education spending.
  - Target education spending to high-labor-demand and high-productivity areas.
- Product market reforms to boost labor demand:
  - Improve the business environment by cutting red tape and opening up regulated professions to facilitate market entry.
  - Promote further integration within the EU to benefit from economies of scale from the single market.
  - Deepen financial markets and improve personal insolvency laws to promote entrepreneurship and innovation.

*Source: IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2018/cr18224.pdf_