## 1sweea2024002

## Source details

**Canonical URL:** [1sweea2024002](https://www.imf.org/-/media/files/publications/cr/2024/english/1sweea2024002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1sweea2024002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1sweea2024002.pdf.json)

---

### Inflation developments
- Inflation in Sweden started rising sharply from June 2022.
- During the global financial crisis, headline inflation (consumer price index with fixed interest rates—CPIF, Box 1) had fallen sharply to just above r one percent by September 2009.
- Inflation rebounded in late 2009, then entered a downward trend in early 2010, staying below the Riksbank’s 2 percent target until 2021Q1.
- Staff projects headline inflation and core inflation to reach target by mid- 2025.
- Sweden’s CPI includes mortgage interest costs; CPIF is CPI with a fixed mortgage interest rate. Authorities publish CPIF excluding energy and/or food as measures of “core” inflation.

### Inflation expectations and surprises
- Inflation has repeatedly surprised forecasters to the upside since the fall of 2021, though these surprises have now normalized (Figure 2: Citi Sweden Inflation Surprise Index).
- Inflation expectations have begun to normalize only since late-2023 and have been stickier for households’ 12-month inflation expectations (Figure 1; Business Tendency Survey: 12 Month Inflation Expectations).
- Market and public expectations affect pricing of financial assets and price- and wage-setting mechanisms.

### Phillips curve analysis and contributions to recent inflation
- Sweden-specific Phillips curve (PC) relates inflation to: past and expected future values of inflation, economic slack, and foreign price developments.
- The model confirms a role for foreign (especially energy and food) price developments in Sweden’s inflation surge since end-2021.
- The model can at most account for 60 percent of the recent surge in inflation, evidenced by sizable positive residuals—especially for core inflation—where actual inflation exceeded model-predicted values in recent quarters.
- Potential factors poorly captured by the PC include:
  - changes in structural relationships (e.g., core inflation becoming more backward-looking),
  - increased pass-through of global commodity prices to domestic inflation after the pandemic,
  - labor shortages,
  - supply bottlenecks,
  - reallocation of demand between goods and services during and after the pandemic,
  - other discretionary policies.

### Risk scenarios and Phillips Curve simulations
- Scenario definitions used in PC simulations:
  - Baseline: WEO assumptions.
  - Negative Supply Shocks: 20 percent rise in energy and food prices.
  - Positive Supply Shocks: 20 percent fall in energy and food prices.
  - Less Slack than Estimated: 2 percentage points lower unemployment gap.
  - More Slack than Estimated: 2 percentage points higher unemployment gap.
  - De-anchoring: 1 percentage point higher expected inflation.
  - More Backward-Looking Price Formation Process: rise in the coefficient on lagged inflation to 0.8 (average of pre-1990s value).
- Key simulation findings:
  - Negative supply shock (20 percent rise in energy and food prices) increases headline and core inflation relative to baseline.
  - Positive supply shock (20 percent fall in energy and food prices) reduces headline and core inflation relative to baseline.
  - A 2 percentage points change in the unemployment gap materially shifts inflation relative to baseline.
  - De-anchoring and increased persistence (more backward-looking formation) raise inflation relative to baseline.
- Shocks have symmetric effects: faster resolution of supply bottlenecks or larger-than-projected slack would accelerate the decline in inflation.

### Profit margins, wages, and inflation drivers (GDP and consumption deflators)
- GDP deflator decomposition components: i) unit profits (gross operating surplus and mixed income per unit of real GDP); ii) unit labor costs (labor compensation per unit of real GDP); iii) taxes (taxes less subsidies per unit of GDP).
- This accounting decomposition shows relative contributions to changes in the GDP deflator (no causal interpretation implied).
- Findings from GDP deflator decomposition:
  - Profits made up half of inflation in Sweden in 2022.
  - Contribution of unit labor cost was very small in 2021 and gradually increased thereafter, reflecting some wage adjustment to the energy shock.
  - In the first half of 2023, contribution of profits to inflation dynamics in Sweden fell slightly, while that of labor costs increased; in the euro area (EA) profit contribution increased over the same period.
  - The role of higher taxes underlying inflation since 2021 may be driven by high ‘revenue buoyancy’.
- Consumption deflator decomposition:
  - Confirms higher share of profits since the pandemic.
  - For 2022–2023H1, Sweden differs markedly from the euro area: Sweden’s consumption deflator inflation is predominantly underpinned by import prices (including exchange rate depreciation effects) and less so by unit profits, unlike in the EA.
  - Contribution of unit labor costs became negative in 2022–2023H1 in Sweden, possibly because of stronger collective bargaining wage setting outcomes.
  - Profit patterns may differ across industries (export oriented vs domestic) and be influenced by business strategy.

### Indicators of inflation momentum (3m/3m by component, Jan-23 to Dec-23)
- Processed food: 13.9, 15.3, 18.5, 16.5, 11.2, 4.3, 0.9, 0.4, 0.5, 0.3, -0.3, 0.5
- Non-energy industrial goods: 8.5, 10.3, 10.9, 10.4, 7.8, 5.4, 4.8, 5.0, 6.1, 4.7, 1.5, -1.8
- Services: 4.4, 4.8, 5.7, 6.6, 7.3, 8.9, 10.0, 8.5, 5.6, 2.8, 1.9, 1.4
- Energy: 7.1, -13.9, -43.6, -44.6, -40.4, -20.7, -25.5, -20.5, -26.7, -15.1, 1.1, 25.3
- Headline: 8.2, 6.7, 4.1, 3.4, 2.5, 3.2, 2.9, 3.2, 2.1, 1.8, 2.3, 2.7

### Exchange rate dynamics — PCA and drivers of krona depreciation
- The Swedish krona has depreciated on a trend basis since 2014; the weakening trend accelerated post-pandemic and has only recently stabilized.
- PCA of nominal effective exchange rates for G-10 cyclical currencies (AUD, CAD, NOK, NZD, SEK) for 2002–2023:
  - The first principal component (PC1) explains about 60 percent of the variance across the five currencies.
  - The second and third components explain about 27 percent and 10 percent of the variance, respectively.
  - PC1 exhibits positive loadings for all five currencies and is strongly correlated with the USD NEER (correlation: -0.85), implying PC1 captures the special role of the USD and global risk-off episodes and monetary policy differences.
  - PC2 is correlated with the EUR NEER (correlation: 0.67).
- Sweden’s de facto monetary policy setting has been relatively accommodative: Sweden’s real ex-post policy rates (deflated with HICP inflation) fell below most G-10 rates during the 2014–2022 period, reflecting subdued domestic inflation and growth outlooks and relative settings compared to the US and euro area.

### Principal components and Nordic-specific factors
- The second principal component has negative loadings for the SEK and the NOK, and positive loadings for the remaining currencies, likely capturing the high correlation between SEK and NOK during the sample period (0.7 on a quarterly basis).
- Possible factors explaining PC2 include:
  - the larger share of trade with the EA of both countries, as confirmed by the high correlation of the component with the EUR nominal effective exchange rate;
  - Norway and Sweden’s large and sustained current account surpluses;
  - market features that make foreign currency synthetic funding in both markets relatively costly (i.e., a breakdown of covered interest parity).

### Deviations from covered interest parity (CIP) and liquidity
- Deviations from CIP became systematic in the Nordics after the global financial crisis, in line with other G-10 currencies.
- In Sweden and Norway the cost of direct funding in USD is lower than the cost of synthetic funding (via FX swaps), as indicated by the negative cross-currency basis.
- Up to 2019, deviations in the Nordics resembled those of defensive currencies rather than those of the cyclical group.
- Contributing factors noted:
  - large current account surpluses in Sweden and Norway that make direct USD funding in local markets relatively more abundant than in other cyclical markets;
  - forward markets in Sweden and Norway have relatively lower liquidity than the rest of the G-10 as measured by the bid-ask spreads on 3-month contracts;
  - other potential factors include regulations that limit financial intermediaries’ risk-bearing capacity and lower liquidity in public debt markets impacting foreign investors’ appetite for SEK-denominated bonds.
- Exact liquidity comparisons reported:
  - On average, the bid-ask spread for the SEK/USD and NOK/USD pairs are 8 bp and 11 bps wider than that of the “Haven” and other cyclical currencies.

### Model-based decomposition of the SEK (EUROMOD / Sweden module)
- During the 2014–21 period, covered interest rate parity shocks (category “Other” in Figure 4) were the main drivers of the nominal exchange rate, followed by monetary policy shocks and demand shocks.
- In the latter part of the sample, Covid-19 related supply and demand shocks played a more important role.
- Monetary policy shocks are defined as deviations of policy rates from the market-based expectations.
- Demand shocks are derived as growth surprises calculated from deviations of actual data from IMF staff expectations one year earlier.
- Supply shocks are derived as inflation surprises calculated from deviations of actual data and IMF staff expectations.

### Third principal component (PC3)
- PC3 has positive loadings for the SEK, NOK and the NZD.
- PC3 appears to capture a cyclical and size-related factor: these currencies share relatively smaller FX markets.
- Wider forward bid-ask spreads point to less liquid markets in Sweden and Norway, signaling a Nordic-specific factor.

### Fiscal and economic implications of LSAPs (Riksbank QE during COVID)
- Context and observed outcomes:
  - The Riksbank’s asset purchases during COVID amounted to some 10 percent of GDP.
  - Those purchases are estimated to have resulted in balance sheet losses of some SEK 61 billion (around 1 percent of 2021 GDP) for the entire period.
  - The Swedish National Audit Office (December 2023) highlighted perceived small effects of LSAPs on economic activity and inflation and recommended against future use of LSAPs with the primary purpose of influencing inflation in Sweden.
- Exercise 1 (conservative empirical estimates):
  - The 10 percent of GDP of QE in Sweden during COVID increased GDP by 1.1 percent relative to a no-QE baseline.
  - QE contributed to 0.5 percent higher inflation at the peak relative to a counterfactual without QE.
  - Assuming Riksbank losses a little over 1 percent of GDP, the QE multiplier (peak output increase divided by the loss of QE) is about unity.
  - A QE multiplier of unity compares favorably to a fiscal spending multiplier typically estimated up to around unity; in severe contractions at the ELB the fiscal spending multiplier may be as high as 2.
- Exercise 2 (macroeconomic model with fiscal dynamics):
  - Under the prevailing outlook at the time of purchases (policy and long-term yields expected to remain low for long), LSAPs:
    - lowered long-term rates and depreciated the exchange rate;
    - helped mitigate the contraction in economic activity and deflationary pressures;
    - created room for an earlier policy lift-off.
  - Had the recovery proceeded as forecasted at the time of purchases, the impact of LSAP on the fiscal stance implied a reduction in government debt by more than 2 percent of annual GDP in the 8-year horizon.
  - This favorable fiscal outcome was mainly due to increased tax revenues and lower debt service costs, which more than offset the adverse impact on central bank profits that cumulate to merely 0.1 percent of annual GDP.
- Scenario with earlier-than-expected monetary tightening:
  - The policy rate needed to be increased much earlier and more sharply than projected due to an unexpected sharp rise in inflation.
  - In this scenario, the cumulative decrease in central bank profits due to LSAPs amounted to a little more that 1 percent of annual GDP.
  - Ex post fiscal gains from LSAPs were significantly reduced under this scenario but still unlikely to be negative overall given the conservative comparison (benefits of QE only in the COVID period versus estimated Riksbank losses for the entire QE period).
- Key quantitative outcomes:
  - Correlation between SEK and NOK during the sample period: 0.7 (quarterly basis).
  - Bid-ask spread differences: SEK/USD and NOK/USD are 8 bp and 11 bps wider than “Haven” and other cyclical currencies.
  - Riksbank QE purchases: some 10 percent of GDP.
  - Estimated Riksbank balance sheet losses for the entire QE period: SEK 61 billion (around 1 percent of 2021 GDP).
  - Estimated GDP increase from QE (conservative): 1.1 percent relative to no-QE baseline.
  - Estimated peak inflation contribution from QE: 0.5 percent higher inflation at the peak.
  - QE multiplier (peak output increase divided by loss of QE): about unity.
  - Modeled reduction in government debt under favorable recovery: more than 2 percent of annual GDP over an 8-year horizon.
  - Cumulated adverse impact on central bank profits in favorable scenario: 0.1 percent of annual GDP.
  - Cumulated decrease in central bank profits in faster-tightening scenario: a little more that 1 percent of annual GDP.
- Policy implications:
  - LSAPs can be a meaningful part of the policy package under exceptional economic circumstances, but risks to central bank balance sheets from earlier-than-expected policy normalization should be considered.
  - Authorities are encouraged to further study:
    - the effectiveness of LSAPs in Sweden;
    - a cost-benefit comparative analysis with other fiscal and monetary policy tools.
  - Considerations highlighted:
    - QE tends to boost private demand and net exports rather than government spending.
    - If a commensurate stimulus had been engineered with fiscal policy, it would likely have been associated with some rise in overall public debt.
    - Exercises do not capture the counterfactual evolution of financial stress or the broader global environment where other major central banks also engaged in QE.

---

### Key findings on pass-through and inflation persistence (section 5)
- Policy rate tightening in the current cycle has resulted in a faster pass-through compared to other AEs, and relative to the past for most lending rates.
- Pass-through for deposit rates is weaker—relative to the past—which should strengthen the effect of monetary tightening on inflation reduction through greater pressure on incomes, though substitution effects between savings and consumption could somewhat weaken this channel.
- Long-term inflation expectations remain well-anchored, while short-term expectations (in particular, household expectations) have drifted up and shown more stickiness, raising the risk that expectation formation becomes more backward-looking and influenced by previous high inflation prints.

### Model, assumptions, and learning mechanism
- Agents update beliefs through adaptive learning: agents form expectations based on a simple statistical model rather than the standard rational expectations assumption.
- The model extends the standard DSGE with expectational learning by Alvarez and Dizioli (2023) and includes:
  - price and wage Phillips curves (relating price and wage inflation to expectations, the gap between real wages and productivity, and economic slack);
  - an IS curve (relating output to the nominal interest rate and inflation expectations);
  - a monetary policy function.
- Heterogeneous agents: a mix of backward- and forward-looking learners with different information sets.
  - Backward-looking learners form expectations based on recent events.
  - Forward-looking learners form expectations rationally using full information, including the share of backward-looking learners. As the share of backward-looking learners increases, forward-looking agents act more like backward-looking agents.
- Near-term inflation expectations and long-term inflation expectations mutually influence each other.

### Channels through which the central bank influences inflation (as modeled)
- Direct demand channel: tighter policy cools demand, lowers the output gap, and hence lowers inflation.
- Expectations channel 1: tightening lowers current inflation that enters agents’ forecasting equations, lowering next-period expectations.
- Expectations channel 2: tightening affects agents’ learning (the coefficients in the forecasting equation); seeing lower-than-expected inflation leads households to update their model of how past inflation matters for future inflation.

### Optimal monetary policy path and staff conditional forecast
- The optimal monetary policy path minimizes a loss function consisting of the output gap and inflation deviations from the 2 percent target, under the assumption that the central bank has full knowledge of current and future shocks and how their actions impact expectations.
- Considering staff’s projections for GDP and inflation up to 2025Q4, and data outturns until 2023Q4, the model’s conditional forecast suggests monetary policy should be tightened further in 2024Q1, before reversing slowly from 2024Q2.
- Optimal interest rate path (quarterly points):
  - 2024Q1: 4.7
  - 2024Q2: 4.4
  - 2024Q3: 3.6
  - 2024Q4: 3.1
  - 2025Q1: 2.5
  - 2025Q2: 2.2
  - 2025Q3: 2.0
  - 2025Q4: 1.8
- Staff recommendation and risk management:
  - The analysis suggests the optimal monetary policy stance would need to be tighter than assumed in the baseline scenario.
  - Given the sharp decline in inflation since October 2023 and an increase in the pace of disinflation since October across most components and core inflation, the MP transmission process (through lower activity) has strengthened.
  - Because of appreciable uncertainty in inflation forecasts, a risk management approach implies maintaining a tight monetary policy stance, with policy rates close to current levels in the first half of 2024 for inflation to return to target by mid-2025.
  - Monetary policy should remain data dependent and nimble to act (tighten or cut) sooner if inflation risks materialize on either side.

### Scenario analysis (summary)
- Upside inflation scenario:
  - Assumes a 10 percent depreciation of the exchange rate within two quarters combined with a proxy for a high-risk premium associated with a drop of GDP in the first quarter.
  - Requires a tighter monetary policy stance.
  - Following the depreciation shock, inflation increases more compared to the baseline and is more persistent, while the output gap becomes more negative. GDP growth recovers faster in growth space but the GDP level remains lower.
- Downside inflation scenario:
  - Focuses on a negative demand shock and requires a looser monetary policy stance.
  - Following the negative demand shock, GDP falls, the output gap turns persistently more negative, and inflation declines relative to the baseline.

### Trade openness and export composition
- Trade openness, as measured by the share of trade in GDP, grew from less than 40 percent of GDP in the early 1990s to 70 percent of GDP in 2022.
- Some three-quarters of exports comprise manufacturing goods (e.g., chemicals and chemical products, machinery and equipment, motor vehicles and electronics) with high import content.
- Main export commodities (percent of total): Industrial machinery 13%, Road vehicles 12%, Electronics, telecommunication 10%, Petroleum products 8%, Pharmaceuticals 7%, Food, beverages, tobacco 6%, Paper, paper products 5%, Other 39%.
- Main export destinations (percent of total): Norway 11%, Germany 10%, USA 9%, Denmark 7%, Finland 7%, United Kingdom 6%, Netherlands 5%, France 4%, Poland 4%, Belgium 4%, China 4%, Other 29%.

### Global value chains, foreign input exposure, and fragility
- Sweden’s trade is diversified across products and markets and integrated along global value chains.
- Foreign Input Reliance (FIR) trends:
  - Sweden’s trade exposure to the US-EUR bloc (including the UK, Germany, and France) has flatlined or declined from 1995 to 2020.
  - Trade exposure to China in manufacturing increased from about 0.8 percent in 1995 to 9 percent in 2020.
  - The share of domestic inputs in manufacturing declined from 77.6 percent in 1995 to 72.7 percent in 2020.
- Fragile intermediate goods:
  - Fragile intermediate goods represent on average 45 percent of total imports by Sweden.
  - Of fragile intermediate imports, 5 percent are supplied by countries from the “Other” bloc, compared to the EU average of 20 percent.
  - Most fragile intermediate goods imported from the “Other” bloc are raw materials (e.g., mineral fuels and oils, aluminum, rare-earth minerals).
  - The external dependence on raw materials for the country’s green transition is even lower.
  - The recent discovery in northern Sweden of one of the largest rare-earth deposits in Europe will reduce vulnerability to supply-side shocks further.

### Policy recommendations and resilience measures
- Continue supply chain diversification, when it helps trade cost efficiency.
- Improve the supply of skilled labor.
- Build strategic reserves.
- Maintain risk assessments and early warning systems for crisis preparedness.

*Source: 1sweea2024002 - 1. Inflation Indicators (PDF chapter), IMF staff calculations and figures as provided in the source content.*

### 1. Inflation Indicators _____________________________________________________________________ 3

### 1. Inflation Indicators

### Inflation developments
- Inflation in Sweden started rising sharply from June 2022.
- During the global financial crisis, headline inflation (consumer price index with fixed interest rates—CPIF, Box 1) had fallen sharply to just above r one percent by September 2009.
- Inflation rebounded in late 2009, then entered a downward trend in early 2010, staying below the Riksbank’s 2 percent target until 2021Q1.
- Staff projects headline inflation and core inflation to reach target by mid- 2025.
- Sweden’s CPI includes mortgage interest costs; CPIF is CPI with a fixed mortgage interest rate. Authorities publish CPIF excluding energy and/or food as measures of “core” inflation.

### Inflation expectations and surprises
- Inflation has repeatedly surprised forecasters to the upside since the fall of 2021, though these surprises have now normalized (Figure 2: Citi Sweden Inflation Surprise Index).
- Inflation expectations have begun to normalize only since late-2023 and have been stickier for households’ 12-month inflation expectations (Figure 1; Business Tendency Survey: 12 Month Inflation Expectations).
- Market and public expectations affect pricing of financial assets and price- and wage-setting mechanisms.

### Phillips curve analysis and contributions to recent inflation
- A Sweden-specific Phillips curve (PC) relates inflation to: past and expected future values of inflation, economic slack, and foreign price developments.
- The model confirms a role for foreign (especially energy and food) price developments in Sweden’s inflation surge since end-2021.
- The model can at most account for 60 percent of the recent surge in inflation, evidenced by sizable positive residuals—especially for core inflation—where actual inflation exceeded model-predicted values in recent quarters.
- Potential factors poorly captured by the PC include:
  - changes in structural relationships (e.g., core inflation becoming more backward-looking),
  - increased pass-through of global commodity prices to domestic inflation after the pandemic,
  - labor shortages,
  - supply bottlenecks,
  - reallocation of demand between goods and services during and after the pandemic,
  - other discretionary policies.

### Risk scenarios and Phillips Curve simulations
- Illustrative risk scenarios show a wide range of possible inflation paths in either direction (Table 1, Figure 4).
- Scenario definitions used in PC simulations:
  - Baseline: WEO assumptions.
  - Negative Supply Shocks: 20 percent rise in energy and food prices.
  - Positive Supply Shocks: 20 percent fall in energy and food prices.
  - Less Slack than Estimated: 2 percentage points lower unemployment gap.
  - More Slack than Estimated: 2 percentage points higher unemployment gap.
  - De-anchoring: 1 percentage point higher expected inflation.
  - More Backward-Looking Price Formation Process: rise in the coefficient on lagged inflation to 0.8 (average of pre-1990s value).
- Key simulation findings:
  - Negative supply shock (20 percent rise in energy and food prices) increases headline and core inflation relative to baseline.
  - Positive supply shock (20 percent fall in energy and food prices) reduces headline and core inflation relative to baseline.
  - A 2 percentage points change in the unemployment gap materially shifts inflation relative to baseline.
  - De-anchoring and increased persistence (more backward-looking formation) raise inflation relative to baseline.
- Shocks have symmetric effects: faster resolution of supply bottlenecks or larger-than-projected slack would accelerate the decline in inflation.

### Profit margins, wages, and inflation drivers (GDP and consumption deflators)
- The GDP deflator is decomposed into: i) unit profits (gross operating surplus and mixed income per unit of real GDP); ii) unit labor costs (labor compensation per unit of real GDP); and iii) taxes (taxes less subsidies per unit of GDP).
- This accounting decomposition shows relative contributions to changes in the GDP deflator (no causal interpretation implied).
- Findings from GDP deflator decomposition:
  - Profits made up half of inflation in Sweden in 2022.
  - Contribution of unit labor cost was very small in 2021 and gradually increased thereafter, reflecting some wage adjustment to the energy shock.
  - In the first half of 2023, contribution of profits to inflation dynamics in Sweden fell slightly, while that of labor costs increased; in the euro area (EA) profit contribution increased over the same period.
  - The role of higher taxes underlying inflation since 2021 may be driven by high ‘revenue buoyancy’.
- Consumption deflator decomposition:
  - Confirms higher share of profits since the pandemic.
  - For 2022–2023H1, Sweden differs markedly from the euro area: Sweden’s consumption deflator inflation is predominantly underpinned by import prices (including exchange rate depreciation effects) and less so by unit profits, unlike in the EA.
  - Contribution of unit labor costs became negative in 2022–2023H1 in Sweden, possibly because of stronger collective bargaining wage setting outcomes.
  - Profit patterns may differ across industries (export oriented vs domestic) and be influenced by business strategy.

*Source: 1sweea2024002 - 1. Inflation Indicators (PDF chapter), IMF staff calculations and figures as provided in the source content.*

### 5.      The previous analysis suggested that sticky core inflation cannot be explained by

### 5.      The previous analysis suggested that sticky core inflation cannot be explained by 

### Key findings on pass-through and inflation persistence
- Policy rate tightening in the current cycle has resulted in a faster pass-through compared to other AEs, and relative to the past for most lending rates.
- Pass-through for deposit rates is weaker—relative to the past—which should strengthen the effect of monetary tightening on inflation reduction through greater pressure on incomes, though substitution effects between savings and consumption could somewhat weaken this channel.
- Long-term inflation expectations remain well-anchored, while short-term expectations (in particular, household expectations) have drifted up and shown more stickiness, raising the risk that expectation formation becomes more backward-looking and influenced by previous high inflation prints.

### Model, assumptions, and learning mechanism
- Agents update beliefs through adaptive learning: agents form expectations based on a simple statistical model rather than the standard rational expectations assumption.
- The model extends the standard DSGE with expectational learning by Alvarez and Dizioli (2023) and includes:
  - price and wage Phillips curves (relating price and wage inflation to expectations, the gap between real wages and productivity, and economic slack);
  - an IS curve (relating output to the nominal interest rate and inflation expectations);
  - a monetary policy function.
- Heterogeneous agents: a mix of backward- and forward-looking learners with different information sets.
  - Backward-looking learners form expectations based on recent events.
  - Forward-looking learners form expectations rationally using full information, including the share of backward-looking learners. As the share of backward-looking learners increases, forward-looking agents act more like backward-looking agents.
- Near-term inflation expectations and long-term inflation expectations mutually influence each other.

### Channels through which the central bank influences inflation (as modeled)
- Direct demand channel: tighter policy cools demand, lowers the output gap, and hence lowers inflation.
- Expectations channel 1: tightening lowers current inflation that enters agents’ forecasting equations, lowering next-period expectations.
- Expectations channel 2: tightening affects agents’ learning (the coefficients in the forecasting equation); seeing lower-than-expected inflation leads households to update their model of how past inflation matters for future inflation.

### Optimal monetary policy path and staff conditional forecast
- The optimal monetary policy path minimizes a loss function consisting of the output gap and inflation deviations from the 2 percent target, under the assumption that the central bank has full knowledge of current and future shocks and how their actions impact expectations.
- Considering staff’s projections for GDP and inflation up to 2025Q4, and data outturns until 2023Q4, the model’s conditional forecast suggests monetary policy should be tightened further in 2024Q1, before reversing slowly from 2024Q2.
- Optimal interest rate path (quarterly points):
  - 2024Q1: 4.7
  - 2024Q2: 4.4
  - 2024Q3: 3.6
  - 2024Q4: 3.1
  - 2025Q1: 2.5
  - 2025Q2: 2.2
  - 2025Q3: 2.0
  - 2025Q4: 1.8

### Staff recommendation and risk management
- The analysis suggests the optimal monetary policy stance would need to be tighter than assumed in the baseline scenario.
- Given the sharp decline in inflation since October 2023 and an increase in the pace of disinflation since October across most components and core inflation, the MP transmission process (through lower activity) has strengthened.
- Because of appreciable uncertainty in inflation forecasts, a risk management approach implies maintaining a tight monetary policy stance, with policy rates close to current levels in the first half of 2024 for inflation to return to target by mid-2025.
- Monetary policy should remain data dependent and nimble to act (tighten or cut) sooner if inflation risks materialize on either side.

### Scenario analysis (summary of effects and policy implications)
- Upside inflation scenario (Figure 4, left column):
  - Assumes a 10 percent depreciation of the exchange rate within two quarters combined with a proxy for a high-risk premium associated with a drop of GDP in the first quarter.
  - Requires a tighter monetary policy stance.
  - Following the depreciation shock, inflation increases more compared to the baseline and is more persistent, while the output gap becomes more negative. GDP growth recovers faster in growth space but the GDP level remains lower.
- Downside inflation scenario (Figure 4, right column):
  - Focuses on a negative demand shock and requires a looser monetary policy stance.
  - Following the negative demand shock, GDP falls, the output gap turns persistently more negative, and inflation declines relative to the baseline.

### Indicators of inflation momentum (Inflation momentum, 3m/3m by component, Jan-23 to Dec-23)
- Processed food: 13.9, 15.3, 18.5, 16.5, 11.2, 4.3, 0.9, 0.4, 0.5, 0.3, -0.3, 0.5
- Non-energy industrial goods: 8.5, 10.3, 10.9, 10.4, 7.8, 5.4, 4.8, 5.0, 6.1, 4.7, 1.5, -1.8
- Services: 4.4, 4.8, 5.7, 6.6, 7.3, 8.9, 10.0, 8.5, 5.6, 2.8, 1.9, 1.4
- Energy: 7.1, -13.9, -43.6, -44.6, -40.4, -20.7, -25.5, -20.5, -26.7, -15.1, 1.1, 25.3
- Headline: 8.2, 6.7, 4.1, 3.4, 2.5, 3.2, 2.9, 3.2, 2.1, 1.8, 2.3, 2.7

### Exchange rate dynamics — principal components analysis (PCA) and drivers of krona depreciation
- The Swedish krona has depreciated on a trend basis since 2014; the weakening trend accelerated post-pandemic and has only recently stabilized.
- PCA of nominal effective exchange rates for G-10 cyclical currencies (AUD, CAD, NOK, NZD, SEK) for 2002–2023:
  - The first principal component (PC1) explains about 60 percent of the variance across the five currencies.
  - The second and third components explain about 27 percent and 10 percent of the variance, respectively.
- PC1 exhibits positive loadings for all five currencies and is strongly correlated with the USD NEER (correlation: -0.85), implying PC1 captures the special role of the USD and global risk-off episodes and monetary policy differences.
- PC2 is correlated with the EUR NEER (correlation: 0.67).
- Sweden’s de facto monetary policy setting has been relatively accommodative: Sweden’s real ex-post policy rates (deflated with HICP inflation) fell below most G-10 rates during the 2014–2022 period, reflecting subdued domestic inflation and growth outlooks and relative settings compared to the US and euro area.

*Source: IMF staff calculations and model results as presented in the provided chapter.*

### 5.      The second principal component likely reflects Nordic-specific factors. The second

### 5.      The second principal component likely reflects Nordic-specific factors. The second

### Principal components and Nordic-specific factors
- The second principal component has negative loadings for the SEK and the NOK, and positive loadings for the remaining currencies, likely capturing the high correlation between SEK and NOK during the sample period (0.7 on a quarterly basis).
- Possible factors explaining PC2 include:
  - the larger share of trade with the EA of both countries, as confirmed by the high correlation of the component with the EUR nominal effective exchange rate;
  - Norway and Sweden’s large and sustained current account surpluses;
  - market features that make foreign currency synthetic funding in both markets relatively costly (i.e., a breakdown of covered interest parity).

### Deviations from covered interest parity (CIP) and Nordic distinctiveness
- Deviations from covered interest parity became systematic in the Nordics after the global financial crisis, in line with other G-10 currencies.
- In Sweden and Norway the cost of direct funding in USD is lower than the cost of synthetic funding (via FX swaps), as indicated by the negative cross-currency basis.
- Up to 2019, deviations in the Nordics resembled those of defensive currencies rather than those of the cyclical group.
- Contributing factors noted:
  - large current account surpluses in Sweden and Norway that make direct USD funding in local markets relatively more abundant than in other cyclical markets;
  - forward markets in Sweden and Norway have relatively lower liquidity than the rest of the G-10 as measured by the bid-ask spreads on 3-month contracts;
  - other potential factors include regulations that limit financial intermediaries’ risk-bearing capacity and lower liquidity in public debt markets impacting foreign investors’ appetite for SEK-denominated bonds.
- Exact liquidity comparisons reported:
  - On average, the bid-ask spread for the SEK/USD and NOK/USD pairs are 8 bp and 11 bps wider than that of the “Haven” and other cyclical currencies.

### Model-based decomposition of the SEK (EUROMOD / Sweden module)
- A model-based shock decomposition using Sweden’s module of the IMF’s EUROMOD system shows:
  - During the 2014–21 period, covered interest rate parity shocks (category “Other” in Figure 4) were the main drivers of the nominal exchange rate, followed by monetary policy shocks and demand shocks.
  - In the latter part of the sample, Covid-19 related supply and demand shocks played a more important role.
- Monetary policy shocks are defined as deviations of policy rates from the market-based expectations.
- Demand shocks are derived as growth surprises calculated from deviations of actual data from IMF staff expectations one year earlier.
- Supply shocks are derived as inflation surprises calculated from deviations of actual data and IMF staff expectations.

### Third principal component (PC3): cyclicality and market size
- The third principal component has positive loadings for the SEK, NOK and the NZD.
- PC3 appears to capture a cyclical and size-related factor: these currencies share relatively smaller FX markets.
- Wider forward bid-ask spreads point to less liquid markets in Sweden and Norway, signaling a Nordic-specific factor.
- Other principal components are not discussed because they mostly represent noise.

### The fiscal and economic implications of LSAPs (Riksbank QE during COVID)
- Context and observed outcomes:
  - The Riksbank’s asset purchases during COVID amounted to some 10 percent of GDP.
  - Those purchases are estimated to have resulted in balance sheet losses of some SEK 61 billion (around 1 percent of 2021 GDP) for the entire period.
  - The Swedish National Audit Office (December 2023) highlighted perceived small effects of LSAPs on economic activity and inflation and recommended against future use of LSAPs with the primary purpose of influencing inflation in Sweden.
- Two analytical exercises performed:
  1. Conservative multiplier estimates from Fabo et al. (2021).
  2. A structural New Keynesian model calibrated to the Swedish economy to estimate economic gains and fiscal cost-benefit.
- Exercise 1 (conservative empirical estimates):
  - The 10 percent of GDP of QE in Sweden during COVID increased GDP by 1.1 percent relative to a no-QE baseline.
  - QE contributed to 0.5 percent higher inflation at the peak relative to a counterfactual without QE.
  - Assuming Riksbank losses a little over 1 percent of GDP, the QE multiplier (peak output increase divided by the loss of QE) is about unity.
  - A QE multiplier of unity compares favorably to a fiscal spending multiplier typically estimated up to around unity; in severe contractions at the ELB the fiscal spending multiplier may be as high as 2.
- Exercise 2 (macroeconomic model with fiscal dynamics):
  - The model is a two-country New Keynesian model with bond market segmentation, augmented for fiscal policy and government debt dynamics, calibrated to initial conditions when Riksbank undertook QE.
  - Under the prevailing outlook at the time of purchases (policy and long-term yields expected to remain low for long), LSAPs:
    - lowered long-term rates and depreciated the exchange rate;
    - helped mitigate the contraction in economic activity and deflationary pressures;
    - created room for an earlier policy lift-off.
  - Had the recovery proceeded as forecasted at the time of purchases, the impact of LSAP on the fiscal stance implied a reduction in government debt by more than 2 percent of annual GDP in the 8-year horizon.
  - This favorable fiscal outcome was mainly due to increased tax revenues and lower debt service costs, which more than offset the adverse impact on central bank profits that cumulate to merely 0.1 percent of annual GDP.
- Scenario with earlier-than-expected monetary tightening:
  - The policy rate needed to be increased much earlier and more sharply than projected due to an unexpected sharp rise in inflation.
  - In this scenario, the cumulative decrease in central bank profits due to LSAPs amounted to a little more that 1 percent of annual GDP.
  - Ex post fiscal gains from LSAPs were significantly reduced under this scenario but still unlikely to be negative overall given the conservative comparison (benefits of QE only in the COVID period versus estimated Riksbank losses for the entire QE period).
  - These outcomes contrast with a fiscal stimulus that, if used to provide a similar boost to private demand and net exports, would likely increase government debt for realistic values of fiscal multipliers.

### Key quantitative outcomes and comparisons (preserve reported figures)
- Correlation between SEK and NOK during the sample period: 0.7 (quarterly basis).
- Bid-ask spread differences: SEK/USD and NOK/USD are 8 bp and 11 bps wider than “Haven” and other cyclical currencies.
- Riksbank QE purchases: some 10 percent of GDP.
- Estimated Riksbank balance sheet losses for the entire QE period: SEK 61 billion (around 1 percent of 2021 GDP).
- Estimated GDP increase from QE (conservative): 1.1 percent relative to no-QE baseline.
- Estimated peak inflation contribution from QE: 0.5 percent higher inflation at the peak.
- QE multiplier (peak output increase divided by loss of QE): about unity.
- Modeled reduction in government debt under favorable recovery: more than 2 percent of annual GDP over an 8-year horizon.
- Cumulated adverse impact on central bank profits in favorable scenario: 0.1 percent of annual GDP.
- Cumulated decrease in central bank profits in faster-tightening scenario: a little more that 1 percent of annual GDP.

### Policy implications and recommendations
- LSAPs can be a meaningful part of the policy package under exceptional economic circumstances, but risks to central bank balance sheets from earlier-than-expected policy normalization should be considered.
- Authorities are encouraged to further study:
  - the effectiveness of LSAPs in Sweden;
  - a cost-benefit comparative analysis with other fiscal and monetary policy tools.
- Considerations highlighted:
  - QE tends to boost private demand and net exports rather than government spending.
  - If a commensurate stimulus had been engineered with fiscal policy, it would likely have been associated with some rise in overall public debt.
  - Exercises do not capture the counterfactual evolution of financial stress or the broader global environment where other major central banks also engaged in QE.

*Source: IMF staff calculations.*

### 1.      Sweden’s economy is very trade oriented. Trade openness, as measured by the share of

### Sweden’s economy is very trade oriented

### Trade openness and export composition
- Trade openness, as measured by the share of trade in GDP, grew from less than 40 percent of GDP in the early 1990s to 70 percent of GDP in 2022.
- Some three-quarters of exports comprise manufacturing goods (e.g., chemicals and chemical products, machinery and equipment, motor vehicles and electronics) with high import content.
- Main export commodities (percent of total): Industrial machinery 13%, Road vehicles 12%, Electronics, telecommunication 10%, Petroleum products 8%, Pharmaceuticals 7%, Food, beverages, tobacco 6%, Paper, paper products 5%, Other 39%.
- Main export destinations (percent of total): Norway 11%, Germany 10%, USA 9%, Denmark 7%, Finland 7%, United Kingdom 6%, Netherlands 5%, France 4%, Poland 4%, Belgium 4%, China 4%, Other 29%.

### Integration in global value chains and foreign input exposure (FIR)
- Sweden’s trade is diversified across products and markets and integrated along global value chains.
- A metric measuring trends in trade, “Foreign Input Reliance (FIR)”, indicates Sweden’s trade exposure to the US-EUR bloc (including the UK, Germany, and France) while still substantial, has flatlined or declined from 1995 to 2020.
- Intra-Nordic trade has on average also fallen.
- Trade exposure to China, particularly in manufacturing (including direct and indirect Chinese inputs in the manufacturing sector), increased from about 0.8 percent in 1995 to 9 percent in 2020.
- The share of domestic inputs in manufacturing declined from 77.6 percent in 1995 to 72.7 percent in 2020.
- Trade restrictions affecting Sweden have steadily risen since the GFC (Global Financial Crisis).

### Dependence on fragile intermediate goods and raw materials
- “Fragile” intermediate goods are identified based on network characteristics of bilateral goods trade (e.g., high centrality of exporters, concentration of imports from a few suppliers, low potential to substitute a supplier).
- Fragile intermediate goods represent on average 45 percent of total imports by Sweden.
- Of fragile intermediate imports, 5 percent are supplied by countries from the “Other” bloc, compared to the EU average of 20 percent.
- Most fragile intermediate goods imported from the “Other” bloc are raw materials (e.g., mineral fuels and oils, aluminum, rare-earth minerals).
- The external dependence on raw materials for the country’s green transition is even lower.
- The recent discovery in northern Sweden of one of the largest rare-earth deposits in Europe will reduce vulnerability to supply-side shocks further.

### Policy recommendations and resilience measures
- Continue supply chain diversification, when it helps trade cost efficiency.
- Improve the supply of skilled labor.
- Build strategic reserves.
- Maintain risk assessments and early warning systems for crisis preparedness.

*Prepared by Alexandra Fotiou, with contributions from Magali Pinat and Reza Yousefi. Sources and notes as provided in the content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1sweea2024002.pdf_