Effects of monetary policy in small open economies during the inflation targeting period

Staff memo Effects of monetary policy in small open econo￾mies during the inflation targeting period Mika Lindgren, Ingvar Strid and Peiyu Wang June 2026

2 Contents 1 Introduction 5 2 Effects of monetary policy in a SVAR model 8 2.1 Model, data and identification 8 2.2 Aggregate effects of monetary policy in the ten SOEs 9 2.3 Country-specific effects of monetary policy 11 2.4 Comparison with other evidence 13 3 Monetary policy transmission elasticities 16 3.1 Okun coefficient 18 3.2 Phillips slope and the sacrifice ratio 18 3.3 Exchange rate pass-through 19 3.4 Statistical significance 22 3.5 What helps explain the heterogeneity in country-variable responses? 23 4 Have the effects of monetary policy changed over time? 26 5 A closer examination of the inflation effects 30 5.1 Difficulties in estimating inflation effects using short-run zero restrictions 30 5.2 Inflation effects with various measures and identification strategies 31 5.3 Potential explanations from the price puzzle literature 33 5.4 Decomposing inflation effects into channels 38 Concluding comments 41 References 42 APPENDIX 49

3 Summary We employ a standardised structural VAR framework to estimate the ef￾fects of monetary policy in ten small open economies (SOEs) with infla￾tion targeting and flexible exchange rates: Australia, Canada, Chile, the Czech Republic, Iceland, New Zealand, Norway, Poland, Sweden and the United Kingdom. While there is a large literature on monetary policy transmission, there remains limited consensus on whether these effects differ across types of economies. Existing evidence on monetary policy transmission in SOEs is often country-specific, while cross-country com￾parisons based on a unified empirical framework are less common. Our comparative analysis contributes by providing directly comparable esti￾mates of the policy effects on inflation, GDP, unemployment, and the ex￾change rate. In the cross-country aggregate, the estimated effects of a policy-rate in￾crease for the group of SOEs align with macroeconomic theory and are consistent with the broader VAR evidence summarised by Enzinger et al. (2025). The estimated magnitudes are also similar to those typically re￾ported for larger economies such as the United States and the euro area in the same study. While there is cross-country variation in the estimated effects, the data do not provide strong statistical support for many of the differences relative to the typical cross-country response. If transmission relationships were similar across countries, one might ex￾pect countries with a large GDP response to a monetary policy shock to also exhibit large unemployment and inflation responses – that is, for variable responses to co-move across countries. We find little evidence of such co-movement. Instead, cross-country differences are more closely associated with monetary policy transmission elasticities, such as the Okun coefficient, the Phillips slope and exchange rate pass-through. For example, larger inflation responses tend to be associated more with a steeper Phillips slope than with a larger GDP response. Furthermore, the policy-conditional Okun and Phillips elasticities appear to be rather well approximated by simpler regression-based estimates. When examining whether the effects have changed over time, two broad conclusions emerge. First, across the ten SOEs, the estimated effects of monetary policy on inflation and GDP appear broadly stable as the esti￾mation sample is extended and, in some cases, increase modestly. Sec￾ond, even where the point estimates suggest an increase, the associated

4 probability intervals generally imply that the changes are not statistically significant in most cases. Lastly, focusing on inflation, we find that estimated responses are partic￾ularly sensitive to both the inflation measure and the identification strat￾egy. In roughly half of the countries, identification based on short-run re￾strictions produces a price puzzle – that is, inflation rises in response to a contractionary policy shock. This issue is mitigated when using core infla￾tion measures that exclude volatile CPI components, or when employing sign restrictions. Decomposing the inflation response further indicates that countries with stronger inflation responses tend to exhibit a larger exchange-rate-related contribution, reflecting both the response of the exchange rate to the policy shock and the pass-through from exchange rates to prices. Authors: Mika Lindgren, Ingvar Strid and Peiyu Wang, Monetary Policy Department, Sveriges Riksbank. 1 1 Many thanks to Vesna Corbo, Stefan Laséen, Henrik Lundvall, and Marianne Nessén for their valuable comments. Peiyu Wang is currently at the National Institute of Economic Research (NIER). The work was conducted while she was at the Monetary Policy Department, Sveriges Riksbank.

5 1 Introduction This paper examines the effects of monetary policy in ten small open economies (SOEs) with inflation targeting regimes and flexible exchange rates: Australia, Canada, Chile, the Czech Republic, Iceland, New Zealand, Norway, Poland, Sweden, and the United Kingdom. Existing evidence on monetary policy transmission in SOEs is often country-specific, and cross-country comparisons based on a unified empirical frame￾work are less common. By estimating comparable structural VAR models across a broad set of SOEs, we provide systematic evidence on how monetary policy affects in￾flation, GDP, unemployment and the exchange rate during the inflation targeting pe￾riod. Since its adoption by the Reserve Bank of New Zealand in 1990, inflation targeting has become the dominant monetary policy framework used by central banks worldwide. It is characterised by (i) an announced numerical inflation target2 , (ii) an implementa￾tion of monetary policy that gives a major role to an inflation forecast, and (iii) a high degree of transparency and accountability.3 The transition to inflation targeting in￾volved abandoning exchange-rate pegs in favour of explicit inflation objectives and a floating exchange rate, which has helped these economies absorb external shocks. Overall, the experience has been favourable: inflation has been lower and more firmly anchored, monetary policy has become more transparent and accountable, and the framework has demonstrated flexibility in responding to large external shocks without eroding credibility.4 Despite its widespread success and a large body of research on monetary policy trans￾mission, there remains limited consensus on whether monetary policy effects differ across types of economies. In particular, relatively few studies analyse multiple non￾euro small open economies in a common standardised framework. Small open econo￾mies differ from large economies primarily in their degree of openness and exposure to external conditions. They rely more heavily on international trade and capital flows, making them more sensitive to global shocks. As a result, exchange rates play a larger role in the transmission of monetary policy, influencing inflation and output more di￾rectly than in larger, more closed economies.5 The central contribution of this staff memo is the analysis of monetary policy trans￾mission elasticities. We examine the joint responses of key macroeconomic variables to a monetary policy shock through an Okun coefficient, a Phillips slope and sacrifice ratio, and exchange rate pass-through. This allows us to assess whether cross-country differences in inflation and unemployment responses mainly reflect differences in GDP and exchange-rate responses, or differences in the elasticities linking these 2 Most central banks conduct what is known as ‘flexible’ inflation targeting, which means that in addition to stabilising inflation, some weight is assigned to stabilising output and employment, see Rogoff (1985). 3 Svensson (2007). 4 Inflation targeting has been successful both in anchoring inflation and managing the large shocks associ￾ated with the Global Financial Crisis and Covid-19 pandemic, see Kiley and Mishkin (2025) and Svensson (2007). For a discussion on how monetary policy has become more transparent with inflation targeting, see Jonsson and Vredin (2025). 5 In contrast, large economies are more driven by domestic demand, less exposed to external develop￾ments, and face weaker exchange rate effects.

6 variables. In addition, the memo contributes by applying a standardised SVAR frame￾work to a broad set of inflation targeting SOEs, yielding directly comparable estimates of the effects of monetary policy on GDP, inflation, unemployment and the exchange rate. Finally, we examine the robustness of the inflation effects across inflation measures and identification strategies, given the well-known sensitivity of inflation re￾sponses in SVAR models. Table 1 documents the adoption of inflation targeting in these countries over the past three decades. The table also shows that these ten economies are deeply integrated into global trade and financial markets, with trade-to-GDP ratios typically ranging be￾tween 50 and 120 percent, and their business cycles are closely linked to develop￾ments abroad. 6 With roughly three decades of inflation targeting experience, this is an appropriate moment to take stock. We implement a standardised structural VAR framework across countries and comparable samples, identifying monetary policy shocks using both short-run zero restrictions and sign restrictions. This approach yields directly comparable quantitative estimates of the effects of monetary policy on inflation, GDP, unemployment, and the exchange rate. Table 1. Ten small open economies with inflation targeting and flexible exchange rates Country Inflation target Flexible exchange rate Trade-to-GDP*, Per cent GDP correlation** Australia 1993 1983 42 0.74 Canada 1991 1970 68 0.89 Czech Republic 1997 1997 122 0.76 Chile 1999 1999 64 0.70 Iceland 2001 2001 81 0.64 Norway 2001 1992 71 0.75 New Zealand 1990 1985 57 0.60 Poland 1998 1999 81 0.72 Sweden 1993 7 1992 85 0.83 United Kingdom 1992 1992 57 0.93 Note. *Historical average, 1995-2024. World Bank. An average for the world equals 54%. **Correla￾tion between year-on-year domestic and foreign GDP growth, using the trade-weighted average of the US and Euro Area. 1996Q1-2024Q4. The first two columns show the year inflation targeting and a flexible exchange rate, respectively, were introduced. The remainder of this staff memo is organised as follows. In Section 2, we present our main results. At the cross-country level, the estimated effects of monetary policy for the group of SOEs align with standard macroeconomic theory, with impulse responses displaying the expected signs. The magnitudes and dynamics of these responses are also in line with the existing literature and do not provide clear evidence that the ef￾fects in small open economies differ systematically from those in larger economies. 6 For example, the correlation between GDP growth in these countries and weighted measures of GDP growth for the euro area and the US range between 0.6 and 0.9 in the period 1996–2024. 7 In Sweden, the inflation target was introduced in 1993 and came into effect in 1995.

7 Although the estimated responses vary across countries, only a few country–variable responses are statistically different from the cross-country median. In Section 3, we examine the joint responses of variables to a monetary policy shock. Across the ten SOEs, we find that cross-country heterogeneity in the effects of mone￾tary policy is better characterised by differences in monetary policy transmission elas￾ticities — the Okun coefficient, the Phillips slope and exchange rate pass-through — than by co-movement among the raw variable responses. If transmission relationships were similar across countries, one might expect countries with a large GDP response to also exhibit large unemployment and inflation responses. We find limited evidence of this. Instead, cross-country differences become clearer when responses are ex￾pressed relative to one another, through transmission elasticities. For example, larger inflation responses tend to be associated more with a steeper monetary policy–condi￾tional Phillips slope than with a larger GDP response. Furthermore, the Okun and Phil￾lips elasticities appear to be well approximated by simpler regression-based esti￾mates. In Section 4, we analyse whether the effects have changed over the inflation targeting period. Two broad conclusions emerge from the expanding-sample analysis. First, across the ten SOEs, the estimated effects of monetary policy on inflation and GDP are broadly stable over time and, in some cases, increase modestly. Second, even where the point estimates suggest an increase, the associated probability intervals generally indicate that these changes are not statistically significant for most country– variable responses. Finally, Section 5 examines inflation responses. We show that the estimated inflation response is sensitive to both the inflation measure and the identification strategy: un￾der short-run zero restrictions, a price puzzle arises in roughly half of the countries. This puzzle largely disappears when using core inflation measures that exclude volatile CPI components or when imposing sign restrictions. A decomposition further indicates that countries with stronger inflation responses tend to exhibit a more pronounced exchange-rate channel.

8 2 Effects of monetary policy in a SVAR model In this section, we present the model, data, and identification approach used to estimate the effects of monetary policy in the ten small open economies (SOEs). We then report the main results, both at the aggre￾gate level and for individual countries. We also compare our findings with the broader SVAR literature and with the available country-specific evidence. At the aggregate level, the estimated effects of a policy-rate increase for the group of SOEs align with macroeconomic theory and are consistent with those reported in the existing literature. Moreover, they do not pro￾vide clear evidence that monetary policy has different effects in SOEs than in larger economies. While there is cross-country variation in the es￾timated effects, the data do not provide strong statistical support for many of the differences relative to the typical (median) response. 2.1 Model, data and identification Model We study the effects of monetary policy in a Bayesian vector autoregressive (VAR) model with a small set of key macroeconomic variables. We estimate the model for ten SOEs, keeping the elements that influence estimation, most importantly the sam￾ple period, model specification (i.e. the variables), prior distribution for the parame￾ters and identification method, as uniform as possible across countries. The model uses quarterly data on eight variables, three foreign and five domestic.8 The foreign variables are trade-weighted measures of log real gross domestic product (GDP), the quarterly change in the consumer price index (CPI), and the policy rate for the United States and the euro area. The domestic variables include log real GDP, the unemployment rate, quarter-to-quarter change in the CPI, a measure of the policy rate, and the log real exchange rate. Thus, all variables enter the model in levels ex￾cept for the foreign and domestic price indices which are included as inflation rates (i.e. first differences). The specification is intentionally simple: we exclude determinis￾tic trends (e.g., linear trends) and dummy variables. We estimate the model over the period 1995Q1–2024Q4 with K = 4 lags and a Minnesota prior on the coefficients. 8 The model specification follows Berggren et al. (2024) closely.

9 Data The data underlying the model are constructed as consistently as possible across the ten small open economies. Harmonised variable definitions enable meaningful cross￾country comparisons of the effects of monetary policy on key macroeconomic varia￾bles. The real exchange rate is constructed in an identical manner for all countries, and the national statistical series for GDP, CPI, and unemployment follow broadly comparable definitions. Consumer price indices are used throughout, with the excep￾tion of Sweden, where the CPI with fixed interest rate (CPIF) is employed. In our nota￾tion, a decline in the real exchange rate corresponds to an appreciation of the domes￾tic currency. The trade-weighted foreign variables are also compiled using a common procedure across countries. Further details on the data sources and the transfor￾mations applied are provided in Table B1 in Appendix B. Identification The monetary policy shock is identified using two standard approaches: short-run zero restrictions and sign restrictions. Under the short-run zero restriction approach, we assume that the shock can affect the nominal, and therefore also the real, ex￾change rate within the same quarter as the change in the policy rate but cannot con￾temporaneously influence the other variables. These assumptions are commonly used in the literature. 9 Domestic monetary policy is further assumed not to affect the for￾eign variables, which is a standard small-open-economy assumption.10 Equivalently, under this identification the central bank’s reaction function in the model allows the policy rate to respond to contemporaneous and lagged values of the foreign variables, to contemporaneous and lagged domestic GDP, unemployment, and inflation, and to lagged values of the policy rate and the real exchange rate. The sign-restriction approach imposes that a contractionary monetary policy shock, which raises the policy rate, leads to an immediate appreciation of the exchange rate and, on average over the subsequent two years, lower GDP and lower inflation. Like￾wise, the average unemployment response is restricted to be positive. In this case we retain the short-run zero restrictions described above and hence the sign restrictions become additional identifying restrictions.11 2.2 Aggregate effects of monetary policy in the ten SOEs Figure 1 reports the median responses across the ten SOEs to an unanticipated tight￾ening that raises the policy rate by 1 percentage point on impact. This normalization is maintained throughout the paper to facilitate comparability. For each country and variable, we first obtain the pointwise posterior median response and then we take 9 Christopher A. Sims first introduced vector autoregressive (VAR) models as an empirical approach to mac￾roeconomic analysis and proposed an identification strategy based on minimal theoretical restrictions. See, for instance, Sims (1980) and Sims (1986). 10 This is achieved by assuming a block exogenous VAR in combination with the short-run zero restrictions. 11 In the sign restriction case, we retain the contemporaneous zero restrictions for all variables except the real exchange rate. A similar hybrid approach is used by e.g. Bjørnland and Halvorsen (2014) who combine zero impact restrictions for GDP and inflation with a sign restriction for the exchange rate. As they discuss, a problem with relying on pure sign restrictions is that the identification scheme will be non-unique.

10 the median across countries to obtain the response shown in the figure.12 Following the shock, the policy rate gradually returns to its baseline over approximately two to three years. Under short-run zero restrictions, the responses broadly align with stand￾ard theory: output falls, unemployment rises, and the real exchange rate appreciates, while the median inflation response is negative but relatively small. Under sign re￾strictions, the qualitative responses are consistent with the imposed restrictions by construction, with output and inflation falling, unemployment rising, and the real ex￾change rate appreciating. The median estimates also imply long-run neutrality of monetary policy, i.e. there are no permanent effects on GDP, unemployment, or the real exchange rate. Under sign-restriction identification, real activity declines gradually and with a delay: real GDP falls by around 0.5 percent after one to two years (median response), while the effect on unemployment peaks somewhat later at about 0.4 percentage points. Hence, real-economy effects are hump-shaped and lagged, and rather persistent. For GDP, the results obtained with (zero) restrictions are broadly similar to those under sign restrictions. Under sign-restriction identification, the nominal, and by implication the real, ex￾change rate responds swiftly, appreciating by roughly 7 percent on impact. This ap￾preciation exceeds what a simple real interest-parity mechanism would predict based solely on the rise in the real policy rate. Annual inflation falls by about 0.3 percentage points, with the maximum effect materialising after one to two years. For both infla￾tion and the exchange rate, the estimated amplitudes under sign restrictions are sub￾stantially larger than with identification based on short-run zero restrictions. In a sub￾set of countries, the short-run restrictions case exhibits a price puzzle, which helps ex￾plain why the median effects are larger (i.e. more negative) under sign restrictions.13 This is discussed further in Section 5. 12 Impulse responses are computed from the posterior distribution of model parameters obtained via the Gibbs sampler. For each country, each posterior draw of the impulse response function (IRF) is first normal￾ized such that the impact response of the policy rate equals a 1 percentage point increase. The median IRF is then computed pointwise across the normalised posterior draws for each country. Finally, the reported response is the pointwise median across these country-specific median IRFs. The aggregate response should therefore be interpreted as a descriptive summary of the typical cross-country response, rather than as the impulse response of any individual country or of an estimated pooled model. See Appendix C for more details. 13 A price puzzle refers to inflation rising (falling) in response to a contractionary (expansionary) policy shock.

11 Figure 1. Median effects of monetary policy in ten small open economies Note. The figure reports the median impulse responses of a 1-percentage-point increase in the policy rate across ten small open economies. The country models are estimated over the period 1995Q1–2024Q4, except for Chile, whose sample begins in 1995Q3. The bands represent second smallest and largest responses across countries and for each quarter. The policy rate, GDP, the unemployment rate and the real exchange rate are measured in levels while inflation is in annual percentage changes. 2.3 Country-specific effects of monetary policy Figure 2 shows the posterior median peak effects of a normalised monetary policy shock in the individual countries, with the horizontal line indicating the median of the country-specific peak effects. In what follows, we emphasise the estimates from the combined identification that includes sign restrictions. These restrictions help ensure that the identified shock aligns with standard monetary transmission mechanisms, while keeping the underlying short-run structure unchanged. However, we note that the relative effects of the different countries are similar with identification based merely on short-run zero restrictions.

12 Figure 2. Peak effects of monetary policy in ten small open economies using sign restriction identification

Note. The figure shows the country median peak impulse responses for four different variables iden￾tified using sign restrictions. GDP, the unemployment rate and the real exchange rate are in levels, whereas inflation is reported in annual percentage changes. The horizontal line refers to the median of the country-specific median peak impulse responses. See Appendix C for more details. We note that the aggregate peak effects implied by Figures 1 and 2 differ somewhat. Figure 2 does not account for uncertainty in the timing of the peak response. As a re￾sult, the peak effects reported there are larger than those inferred from Figure 1. These differences are discussed in detail in Appendix C. For most countries, the peak effects on inflation range from −0.2 to −0.6 percentage points, while Poland and Sweden show significantly larger declines, around −1 per￾centage point. The peak effects on GDP lie in the range of −0.6 to −0.9 per cent, with Norway and New Zealand exhibiting larger declines and Chile showing a relatively small response. The peak effects on unemployment generally fall between 0.3 and 0.6 percentage points, with Chile again showing a comparatively smaller effect. The peak responses of the real exchange rate show a wide dispersion. Chile’s response is once more relatively small, whereas Australia, Canada and New Zealand exhibit compara￾tively larger appreciations.14 Across all variables, the effects of monetary policy in Chile are relatively small, below the median for every variable, while the effects are comparatively large in Canada and Sweden. The Czech Republic and the UK display effects that are close to the overall median across variables. New Zealand and Norway, in turn, are characterised by 14 Our estimated effects on the exchange rate are well in line with those reported by Bjørnland and Hal￾vorsen (2014) based on SVAR models estimated for a subset of the countries studied here in the period 1990-2009. They find relatively large effects for Australia and New Zealand, average effects for Sweden and UK, and relatively small effects for Norway.

13 relatively large effects on GDP but more moderate responses in unemployment and inflation. Finally, Table 2 evaluates the statistical significance of the cross-country differences highlighted in Figure 2. Specifically, for each country and variable we report the poste￾rior probability that the country-specific peak effect exceeds the cross-country me￾dian peak effect (depicted by the black line in Figure 2). A probability close to one in￾dicates that the country-specific peak effect is likely to be above the cross-country median, while a probability close to zero indicates that it is likely to be below the me￾dian. We interpret probabilities above 90 per cent, or below 10 per cent, as indicating a statistically meaningful deviation from the median; these entries are highlighted in the table. While the estimated peak effects vary across countries in ways that may be economically relevant, the probabilities suggest that a substantial share of these dif￾ferences cannot be distinguished from the cross-country median with high statistical confidence under this criterion. Notably, Chile exhibits statistically significantly smaller effects than the cross-country median across all variables. In addition, the GDP response in Norway is significantly larger than the median, and the inflation re￾sponse in Sweden is significantly larger than the median. Table 2. Probability that the peak effect is larger than the cross-country median peak effect GDP Unemployment Real exchange rate Inflation Australia 0.44 0.26 0.73 0.65 Canada 0.55 0.73 0.65 0.45 Chile 0.05 0.01 0.01 0.10 Czech Republic 0.35 0.41 0.40 0.53 Iceland 0.52 0.50 0.70 0.53 Norway 0.94 0.49 0.33 0.24 New Zealand 0.80 0.42 0.61 0.35 Poland 0.34 0.61 0.32 0.82 Sweden 0.55 0.59 0.56 0.90 United Kingdom 0.47 0.66 0.45 0.30 Note. The table shows the probability that the peak effect on the variable for a given country is larger than the cross-country median peak effect which is shown in Figure 2. 2.4 Comparison with other evidence Table 3 compares our cross-country aggregate (median) results for the ten SOEs with three external benchmarks: the Macroeconomic Model Database (MMD), the meta￾study by Enzinger et al. (2025) and a selection of country-specific studies for our group of SOEs. This comparison yields two main conclusions. First, our estimated ef￾fects are broadly consistent with those reported in the wider research literature. Sec￾ond, the estimated effects of monetary policy in SOEs are of a similar order of

14 magnitude to those reported for larger economies such as the United States or the euro area. Table 3. Effects of monetary policy in a large number of models and studies Models and studies Peak effects on GDP Peak effects on inflation Small open economies, 10 countries (This paper) Short-run zero restrictions (Figure D3) -0.5 -0.2 Sign restrictions (Figure 2) -0.8 -0.5 Macroeconomic Model Database United States, 48 models -0.5 -0.2 Euro area, 16 models -0.5 -0.3 Meta analysis by Enzinger et al. (2025) All (bias-corrected), 409 studies -0.3 -0.2 Short-run zero restrictions -0.5 -0.2 Sign restrictions -0.9 -0.5 High-frequency -1.8 -0.4 Narrative -0.5 -0.5 United States -0.6 -0.3 Euro area -0.6 -0.2 Other advanced economies -0.4 -0.2 Emerging economies -0.4 -0.2 Selected country-specific studies, Median of peak effects Australia -0.8 -0.4 Canada -1 -0.5 Chile -2 -0.8 Czech Republic - -0.4 Iceland -0.9 -0.6 Norway -1 -0.3 New Zealand -0.2 -0.8 Poland -0.7 -0.4 Sweden -0.8 -0.4 United Kingdom -1 -0.6 Median of country-medians -0.9 -0.5 Note. The effects reported in the table refer to GDP in levels and inflation measured as the annual percentage change. The policy rate effect is normalised to 1 percentage point. For the selected country-specific studies the median for each country is calculated based on studies employing VAR models with short-run zero or sign restriction identification methods. See Table 6 in Section 5.3 for study-specific results underlying the median for each country. Using the MMD, we extract impulse responses from a broad set of Dynamic Stochastic General Equilibrium (DSGE) models estimated on U.S. and euro-area data. In terms of

15 peak effects, the benchmarks from these models are of similar magnitude to our SOE estimates for both GDP and inflation. Turning to the meta-study, Enzinger et al. (2025) assemble a large set of published im￾pulse responses and allow comparisons by identification strategy and country group￾ing. Our SOE sample aligns most closely with their “other advanced” economies cate￾gory.15 Both under short-run zero restrictions and sign restrictions, our estimated ef￾fects are broadly in line with the corresponding estimates in the meta-study. The comparison with high-frequency identification is also informative. Our GDP effect is smaller than that obtained in high-frequency studies, while the inflation effect is similar. This mirrors the broader pattern documented by Enzinger et al. (2025), where high-frequency studies imply substantially larger GDP effects than other identification approaches, while inflation effects differ less.16 Our results are broadly in line with the country and regional benchmarks summarized in Table 3, and in particular with the “other advanced” economies that most closely match our set of SOEs. To further compare our results with studies focusing specifi￾cally on the small open economies included in our sample, we also benchmark our ag￾gregate estimates against a range of country-specific studies.17 Overall, our results ap￾pear broadly consistent with the existing literature. We obtain similar effects on GDP: around –0.8 compared with a median of –0.9 in the selected country studies, as well as comparable inflation responses (–0.5 compared with a median of –0.5). The coun￾try-level medians are drawn from VAR models using either short-run zero or sign re￾strictions; however, we only have a few relevant studies for each country and both model specifications and empirical results differ substantially across studies, and, fur￾thermore, the results are often sensitive to the choice of inflation measure. For these reasons, the comparisons should be interpreted with caution and seen primarily as in￾dicative of the general magnitude of monetary policy effects rather than precise quantitative differences. Table 6 in Section 5.3 summarises the underlying country￾specific studies, and Section 5 discusses the robustness of our estimated inflation re￾sponses when alternative inflation measures are used. 15 In Enzinger et al. (2025), the “other advanced countries” category includes, for example, Australia, Can￾ada, Japan and the UK. 16 For example, in the case of Sweden, Almerud et al. (2024) use a high-frequency approach and obtain rela￾tively large effects of a policy shock on GDP. 17 The country-specific studies represent a selected sample, and the number of available studies varies across countries. Our focus is primarily on studies employing VAR models with short-run zero or sign re￾striction identification and sometimes conducted in close association with the respective national central banks. These studies are used to compute the country-specific median estimates in Table 3, since they are closest to our empirical specification. However, we also report evidence from studies using high-frequency, external-instrument, narrative and FAVAR approaches when available, as these provide important bench￾marks from the recent empirical literature. These estimates are not included in the medians because their coverage across countries is limited and because the identifying variation differs from that in our baseline SVAR framework. In some cases, results from studies based on DSGE (dynamic stochastic general equilib￾rium) models are also reported, although not included in the median values reported for each country. Fur￾ther details on the selected country-specific studies are provided in Table 6 in Section 5.3.

16 3 Monetary policy transmission elasticities In this section, we examine whether there are systematic relationships in how key macroeconomic variables respond to a monetary policy tighten￾ing across the ten SOEs. For example, do countries with relatively large real-economy responses also tend to exhibit larger inflation responses? To address this question, we compare monetary policy transmission elas￾ticities, defined as summary measures that capture the joint responses of GDP, unemployment, inflation, and the exchange rate to a monetary pol￾icy shock. We also assess whether simple regression-based estimates of these elasticities, which are often used as benchmark relationships in in￾formal policy discussions, are consistent with the monetary-policy-condi￾tional estimates implied by the VAR models. If transmission relationships were similar across countries, one might ex￾pect countries with a large GDP response to a monetary policy shock to also exhibit large unemployment and inflation responses. Across the ten SOEs, we find little evidence of such systematic co-movement among the raw variable responses. Instead, cross-country heterogeneity in the ef￾fects of monetary policy is better characterised by differences in mone￾tary policy transmission elasticities, such as the Okun coefficient, the Phillips slope and exchange rate pass-through. For example, larger infla￾tion responses tend to be associated more with a steeper Phillips slope than with a larger GDP response. Furthermore, the policy-conditional Okun and Phillips elasticities appear to be rather well approximated by simpler regression-based estimates. To characterise cross-country differences in the relative magnitude of the responses to a monetary policy shock, we compute transmission elasticities defined over pairs of variables. We focus on three sets of elasticities: an Okun-type elasticity linking GDP and unemployment; a Phillips-type elasticity and the associated sacrifice ratio linking GDP and inflation; and exchange rate pass-through (ERPT) linking the exchange rate and inflation.18 We also compare these monetary-policy-conditional (MP-conditional) elasticities with conventional regression-based (unconditional) estimates of the same elasticities. Sim￾ple OLS regressions are frequently used in practical policy work to form benchmark 18 We use the term transmission elasticities to distinguish these objects from the monetary policy multipli￾ers studied by, for example, Barnichon and Mesters (2021) and Alessandri, Jordà and Venditti (2025). Those papers use multiplier concepts to characterise the effectiveness or trade-offs of monetary policy through cumulative response ratios. Our measures are instead MP-conditional versions of standard structural rela￾tionships — Okun’s law, the Phillips curve and exchange rate pass-through — computed from the impulse responses to an identified monetary policy shock.

17 views on these relationships. It is therefore useful to assess whether such approxima￾tions are potentially misleading in this context, or whether they provide a reasonable guide despite not conditioning on a monetary policy shock. The regression specifica￾tions are described in detail in Appendix E. Under Okun’s law, assuming a common Okun coefficient, a monetary policy tightening that induces a larger decline in GDP in one country than in another should also be ac￾companied by a larger rise in unemployment, implying broadly similar transmission from output to labour market conditions across countries. Likewise, consistent with the Phillips-curve relationship and standard exchange rate pass-through theory, larger GDP contractions and stronger exchange rate appreciations should be associated with larger declines in inflation. If the responses of GDP, unemployment, the exchange rate, and inflation line up in this manner across countries, this would suggest that the relationships between these variables are broadly similar, even though the responses of GDP and the exchange rate to the monetary policy shock itself may differ across countries. Conversely, sizeable differences in inflation responses despite comparable GDP or exchange rate movements would point to cross-country heterogeneity in the Phillips slope and/or exchange rate pass-through. There are several ways to construct MP-conditional elasticities. A natural approach is to base them on peak responses, which highlight the largest effects of a monetary policy shock and, importantly, deliver a single scalar elasticity per country that is straightforward to compare across countries. Peak-based measures also avoid placing weight on long horizons where estimated impulse responses may be noisy and impre￾cise. The drawback is that peaks may occur at different horizons across variables, cre￾ating a timing mismatch, and ratios of peaks need not coincide with ratios computed at a common horizon. An alternative is to use average or cumulative responses over a given horizon window, which aligns timing more explicitly and smooths out small ir￾regularities in the impulse responses, but at the cost of introducing a horizon choice and potentially generating multiple elasticity measures per country. In this paper, we compute the MP-conditional Okun coefficient and Phillips slope us￾ing peak responses of unemployment, GDP, and inflation, as this captures the largest effects of a monetary policy shock while facilitating cross-country comparisons across our ten SOEs. As a robustness check, we also consider cumulative (or window-aver￾age) versions of these elasticities and find that the qualitative patterns and cross￾country comparisons are generally similar. That is, our main conclusions regarding the Okun and Phillips relationships would not be materially affected if cumulative or aver￾age responses were used instead of peaks. Figure 3 reports the estimated MP-conditional elasticities and Figure 4 relates the MP￾conditional elasticities to corresponding regression-based (unconditional) estimates of the elasticities. Figure 5 presents the relationships between variable responses and MP-conditional elasticities. Throughout this section, the MP-conditional elasticities are computed from the base￾line impulse responses identified using the combined short-run zero and sign re￾strictions. They should therefore be interpreted as conditional on this identification

18 scheme. The sign restrictions impose the qualitative direction of the responses of GDP, inflation, unemployment and the exchange rate, but they do not mechanically determine the relative magnitudes of these responses or the cross-country ranking of the elasticities. Still, the estimated heterogeneity in the elasticities should be viewed as reflecting the data together with the identifying assumptions. 3.1 Okun coefficient The Okun coefficient measures the elasticity that relates changes in the unemploy￾ment rate to changes in real activity. We estimate the MP-conditional Okun coeffi￾cient as the ratio of the peak responses of the unemployment rate and real GDP to a monetary policy shock. We compare this with an unconditional estimate derived from a simple dynamic specification of the Okun relationship.19 The estimates are displayed in Figure 3. Across countries, the median Okun coefficient conditional on a monetary policy shock is 0.6, shown by the black solid line in Figure 3, whereas the median un￾conditional estimate is somewhat lower at 0.4. 20 Both figures are broadly consistent with the ≈0.5 typically reported for unconditional Okun’s law estimates in advanced economies. 21 There is meaningful cross-country heterogeneity in the MP-conditional estimates: Norway and New Zealand display relatively low values at around 0.3, whereas the es￾timates of Canada, Poland and the UK are relatively large, indicating a comparatively strong unemployment response relative to output. For the remaining countries the estimates are typically rather close to the median across countries. Figure 4 further displays a positive relationship between the country-level uncondi￾tional and MP-conditional coefficients: economies in which unemployment reacts strongly to output changes in general (i.e. for different types of shocks) also exhibit a strong unemployment response to MP-induced output changes. The MP-conditional estimates are larger for slightly more than half of the countries, while for the remain￾der they are roughly similar to the unconditional estimates. In summary, this result suggests that the MP-conditional estimate of the Okun coefficient may be reasonably well approximated by a simple unconditional estimate. 3.2 Phillips slope and the sacrifice ratio The Phillips slope links inflation to economic slack, while the sacrifice ratio measures the cumulative real-activity cost of disinflation; the two quantities are naturally in￾versely related. We calculate the MP-conditional Phillips slope as the ratio of the peak 19 See Equation (1) in Appendix E. 20 Our results remain robust when the coefficient is instead calculated as the ratio of cumulative, or, equiva￾lently, average effects on unemployment and GDP over different horizons; see, for example, Ziegenbein (2021) for this approach. The median Okun coefficient across countries computed in this way is 0.4, 0.5 and 0.5 at the horizons 8, 12 and 20 quarters, see Table E1 in Appendix E. 21 See, e.g., Ball, Leigh, and Loungani (2013). Our median estimate is larger than the MP-specific estimates for the United States reported by Daly et al. (2014) and Ziegenbein (2021), and larger than the demand-spe￾cific estimate in Foroni and Furlanetto (2022) for the United States and the euro area; those papers place the shock-specific Okun elasticity around 0.3.

19 responses of inflation and real GDP to a positive policy-rate shock. The cross-country median slope is 0.7 (Figure 3). 22 For comparison, the median MP-conditional Phillips slope in the macroeconomic model database is 0.4 for models estimated using U.S. data and 0.6 for those estimated using euro-area data.23 There is significant country heterogeneity: the slope is larger for Poland and Sweden, while it is comparatively small for Norway, New Zealand and the UK. For the remain￾ing countries the estimated slope is reasonably close to the cross-country median esti￾mate. We calculate the sacrifice ratio as the ratio of cumulative GDP losses to the price-level effect over three years.24 The pattern of the sacrifice ratio is broadly consistent with that of the Phillips slope discussed above: it is lower in Poland and Sweden, where the Phillips slope is larger and policy-induced disinflation is relatively strong, and higher in Norway, New Zealand and the UK, where the Phillips slope is smaller. The median sac￾rifice ratio is approximately 1.7, which broadly aligns with unconditional episode￾based estimates for advanced economies. 25 Our estimates are closely aligned with the median MP-conditional sacrifice ratios in the macroeconomic model database: 2.1 for models estimated using U.S. data and 1.6 for those estimated using euro-area data. This implies that there is no clear evidence that the MP-conditional Phillips curve slope or sacrifice ratio differs between small open economies and large economies. When comparing our MP-conditional and unconditional estimates of the Phillips slope, we find a positive relationship between the two, see Figure 4. 26 For most econ￾omies, unconditional estimates are either as large as or larger than the MP-condi￾tional ones. Only Iceland and the UK exhibit a stronger MP-conditional Phillips slope compared to their unconditional counterpart. Similar to the case of the Okun coeffi￾cient, this comparison exercise suggests that the MP-conditional Phillips slope may be rather well approximated by a simple regression-based estimate of the slope. 3.3 Exchange rate pass-through The exchange rate pass-through (ERPT) is calculated as the ratio of the effect on the price level after three years to the peak effect on the nominal exchange rate. 27 The 22 If the Phillips slope is instead computed as the ratio of the average effects on inflation and GDP, the me￾dian across countries equals 0.6, 0.6, and 0.5 at horizons of 8, 12, and 20 quarters, respectively, see Table E1 in Appendix E. 23 We compute the Phillips slope and the sacrifice ratio for the models in the database ourselves which guarantees that an identical definition is applied. 24 The median sacrifice ratio, as well as cross-country differences in magnitude, remains robust when com￾puted using two- or five-year horizons. When defined this way, the ratio corresponds to the cumulative GDP loss required to reduce the price level by one percent, expressed in percentage points of one year of annualised GDP. 25 Mazumder (2014) reports a median sacrifice ratio of 2.0 for 78 disinflation episodes in OECD countries in the period 1974 to 2003. For a subset of 16 “intentional disinflations”, i.e. disinflations caused by “pure de￾mand shocks”, the median sacrifice ratio is 2.8, see Acevedo and Hofstetter (2023). Tetlow (2022) reports sacrifice-ratio estimates from historical episodes ranging from 0 to 18, based on a selection of studies con￾ducted on the U.S. 26 See Equation (2) in Appendix E for the regression specification for the unconditional Phillips slope. 27 For most countries the peak effect on the nominal exchange rate occurs instantaneously, i.e. in the quar￾ter when the policy rate is changed.

20 cross-country median MP-conditional ERPT is approximately 0.1, meaning that a mon￾etary policy shock which appreciates the nominal exchange rate by 1 percent leads to a 0.1 percent decrease in the price level after three years. The corresponding effects on the price level at the 2- and 5-year horizons are 0.1 and 0.2 percent, respectively.28 There is considerable heterogeneity in the ERPT, with estimates ranging from 0.05 to 0.4 at the three-year horizon across countries. The ERPT appears relatively high in Chile and Poland, while it is relatively small in Iceland and the UK.29 Figure 4 shows that there is no clear relationship between the unconditional and MP-conditional ERPT.30 Therefore, unlike for the Okun coefficient and the Phillips slope, the uncondi￾tional estimates do not provide a useful approximation to the MP-conditional ERPT. This may reflect that exchange-rate movements in the data are driven by several dif￾ferent shocks, and that pass-through to prices can differ depending on the source of the exchange-rate movement. 28 See Table E1 in Appendix E. Our MP-conditional estimate of long-run ERPT for Sweden is consistent with that reported by Corbo and Di Casola (2022), who report a long-run ERPT in response to a monetary policy rate shock to be 0.2 using their benchmark model. 29 The relative discrepancy in ERPT between advanced and emerging economies is supported by Jašová et al. (2016), who found that different pass-through measures remain relatively stable at lower levels in ad￾vanced economies compared to emerging economies, such as Chile and Poland. 30 Our unconditional estimates of long-run ERPT for Australia, Canada and the UK are consistent with those reported by Anderl and Caporale using the same linear specification (2023), although there are some differ￾ences for New Zealand and Sweden. Anderl and Caporale (2023) found that the unconditional long-run ERPT is 0.49 for the U.S. and 0.39 for the euro area.

21 Figure 3. Monetary policy transmission elasticities for ten SOEs Note. Okun coefficient is calculated as the ratio of the peak effect on the unemployment rate to the peak effect on GDP. Phillips slope is calculated as the ratio of peak inflation response to the peak ef￾fect on GDP. Sacrifice ratio is calculated as the cumulative GDP loss divided by the cumulative infla￾tion response, i.e. price level, over a 3-year horizon. Exchange rate pass-through is calculated as the cumulative inflation response divided by the peak nominal exchange rate over a 3-year horizon. The horizontal line refers to the median of the country-specific elasticities.

22 Figure 4. Monetary policy transmission elasticities and regression estimates for ten SOEs Note: This figure reports the MP-conditional and unconditional Okun coefficient, Phillips slope and exchange rate pass-through. The unconditional elasticities are estimated with regressions and data outlined in Appendix E. The country-specific MP-conditional elasticities shown are identical to those reported in Figure 3. The black dashed lines represent the 45-degree line. 3.4 Statistical significance Table 4 evaluates the statistical significance of the cross-country differences high￾lighted in Figure 3. Specifically, for each country and monetary policy elasticity, we re￾port the posterior probability that the country-specific ratio exceeds the cross-country median ratio (depicted by the black line in Figure 3). A probability close to one indi￾cates that the country-specific elasticity is likely to be above the cross-country me￾dian, while a probability close to zero indicates that it is likely to be below the median. We interpret probabilities above 90 per cent, or below 10 per cent, as indicating a sta￾tistically meaningful deviation from the median; these entries are highlighted in the table. While the estimated ratios vary across countries in ways that may be economi￾cally relevant, the probabilities suggest that a substantial share of these differences

23 cannot be distinguished from the median with high statistical confidence under this criterion. Notably, Poland displays a significantly larger Phillips elasticity and exchange rate pass-through than the cross-country median, while its Okun coefficient is close to but below the significance threshold. In addition, the Phillips elasticity in Norway and New Zealand is significantly smaller than the median, and exchange rate pass-through in the Czech Republic is significantly larger than the median. Table 4. Probability that the monetary policy transmission elasticity exceeds the cross-country median Okun Phillips Sacrifice Ratio Exchange Rate Pass-Through Australia 0.29 0.59 0.31 0.49 Canada 0.67 0.34 0.55 0.28 Chile 0.36 0.56 0.42 0.89 Czech Republic 0.60 0.61 0.43 0.92 Iceland 0.48 0.42 0.66 0.17 Norway 0.03 0.01 0.99 0.50 New Zealand 0.12 0.09 0.83 0.34 Poland 0.89 0.92 0.10 0.98 Sweden 0.51 0.77 0.33 0.86 United Kingdom 0.65 0.32 0.63 0.33 3.5 What helps explain the heterogeneity in country￾variable responses? In Section 2.3 we compared the effects of monetary policy across the ten SOEs. Alt￾hough the estimated impulse responses display some cross-country heterogeneity, relatively few country–variable responses are statistically distinguishable from the cross-country median. In this section, we instead focus on monetary policy elasticities, that is, ratios of responses that summarize key trade-offs in the transmission mecha￾nism. These elasticities also exhibit heterogeneity, but once again only a minority are estimated with sufficient precision to be statistically significant. Figure 5 sheds light on the sources of cross-country heterogeneity by relating each country’s estimated variable responses to the corresponding MP-conditional elastici￾ties. Specifically, for each pair of variables (e.g., GDP and inflation), the figure reports two scatter plots: one plotting the response of each variable against the response of the other, and another plotting the response of a given variable against the associated elasticity (e.g., the Phillips slope), computed as the relevant ratio of responses. This comparison helps separate heterogeneity arising from differences in the magnitude of underlying responses from heterogeneity reflecting differences in the transmission trade-offs summarized by the elasticities.

24 Given that the comparison is based on only ten countries, Figure 5 should be inter￾preted as descriptive evidence rather than as a formal statistical test of the cross￾country relationships. Individual countries can have a visible influence on the fitted re￾lationships, and Chile in particular is an outlier in several dimensions. We therefore view the figure as illustrating broad patterns in the data. The main qualitative mes￾sage is that inflation and unemployment responses appear more closely related to the corresponding transmission elasticities than to the underlying GDP or exchange-rate responses themselves. Consider first the relationship between output and inflation. The upper-right panel shows a weak association between countries’ GDP and inflation responses, suggesting that cross-country differences in the inflation response cannot be explained simply by differences in the output response. By contrast, the upper-left panel reveals a notice￾ably tighter relationship between the inflation response and the Phillips slope. This pattern indicates that heterogeneity in the MP-conditional Phillips trade-off accounts for an important share of the cross-country variation in inflation responses. A similar conclusion emerges for unemployment. The scatter plot relating GDP and unemployment responses displays little systematic relationship across countries. In￾stead, cross-country differences in unemployment responses appear more closely aligned with variation in the Okun elasticity, suggesting that heterogeneity in the rela￾tionship between output and unemployment plays a central role in shaping labour market outcomes following a monetary policy shock. Finally, the association between exchange rate responses and price-level responses is also weak, whereas the link be￾tween inflation (or the price level) and the exchange rate pass-through is compara￾tively stronger. This again points to heterogeneity in exchange rate pass-through, ra￾ther than differences in the size of the exchange-rate response itself, as the more im￾portant source of cross-country differences in inflation effects. Taken together, the patterns in Figure 5 should be interpreted cautiously. The correla￾tions are not uniformly strong, and with only ten countries the fitted relationships are sensitive to individual observations. Nevertheless, the figure suggests that cross-coun￾try variation in transmission elasticities helps account for differences in variable re￾sponses, whereas relationships between pairs of raw variable responses are generally weaker. Put differently, we find limited evidence that countries with larger real re￾sponses necessarily exhibit larger inflation responses; instead, countries with larger MP-conditional Phillips slopes tend to display larger inflation responses.

25 Figure 5. Peak impulse responses in key variables and monetary policy transmission elasticities for ten SOEs Note. The left column shows subplots relating peak impulse responses of the variables to MP￾conditional elasticities. The right column shows subplots relating the components of the MP￾conditional elasticity to one another, i.e., the peak impulse responses from which the elasticity is constructed. The units of the responses for inflation, GDP, unemployment, price level and nominal exchange rate are the same as those in Figure 1. The black dashed line denotes the fitted trend line.

26 4 Have the effects of monetary policy changed over time? In this section, we estimate the SVAR model over alternative sample peri￾ods to assess whether the effects of monetary policy on inflation and GDP have changed over the inflation targeting era. This also allows us to examine whether extreme episodes, most notably the COVID-19 pan￾demic, have influenced the peak effects estimated in the full sample. Two conclusions emerge from the results. The estimated effects of mon￾etary policy on inflation and GDP across the ten SOEs are broadly stable across expanding samples and, in some cases, have increased modestly. Moreover, where the estimates do suggest an increase, the associated probability intervals generally indicate that these changes are not statis￾tically significant in most cases. To examine whether and how the effects of monetary policy have changed as the sample is extended, we estimate the model under sign-restriction identification for different sub-periods within the sample spanning 1995Q1 to 2024Q4.31 The purpose of this approach is to obtain estimates that rely on varying degrees on older versus more recent data within the inflation targeting regime. This also makes it possible to assess to what extent periods characterised by extreme events, such as the pandemic years, influence the estimated effects in the full sample. We estimate the SVAR model using short-run zero and sign restrictions on an expanding sample with a fixed starting quarter and a varying end quarter. The initial sample covers the period from 1995Q1 to 2009Q4, while the final sample extends from 1995Q1 to 2024Q4. By compiling the estimates from all expanding samples, we obtain peak effect estimates that are based on different sample periods, allowing us to assess how the estimated effects of mone￾tary policy on GDP and inflation have changed as the sample has been extended. The analysis focuses on the magnitude of the peak effects and does not examine whether the timing of monetary policy transmission has changed over time. Figure 6 illustrates, for each SOE, how the estimated peak effect of a monetary policy shock on inflation changes over time, where the point on the x-axis denotes the sam￾ple endpoint.32 Figure 7 shows the corresponding peak effect on GDP. In all cases, the responses are normalised to an unexpected one-percentage-point increase in the 31 The estimation follows the model and identification outlined in Section 2. The expanding-sample ap￾proach should be interpreted as a sequential updating exercise: it shows how the estimated effects change as additional observations become available, while keeping the early part of the inflation targeting period in the estimation sample. This provides a natural Bayesian interpretation, as the posterior estimates are up￾dated when the information set is expanded. However, it is not designed to identify structural breaks at specific dates. Rolling-window or time-varying-parameter approaches could be more sensitive to such changes but would also rely on shorter effective samples and typically imply less precise estimates. 32 We were not able to estimate the effects for Iceland using the expanding sample due to computational constraints. As a result, Iceland is excluded from the time-varying analysis based on expanding samples.

27 policy rate in the initial period.33 The figures also display probability intervals for the estimated peak effects and their evolution over time; for example, the outer bands represent a 90 per cent probability interval. These intervals illustrate the uncertainty surrounding the estimates and provide an informal basis for assessing whether the es￾timated effects have changed significantly over time. We draw two overall conclusions from the results. First, the estimated effects of mon￾etary policy on GDP and inflation in the ten countries are broadly stable over time and, in some cases, have increased somewhat. Second, where the estimates do indi￾cate an increase, the associated probability intervals generally imply that the change is not statistically significant in most cases. To assess more formally whether the estimated effects have changed over time, we compare results for the samples 1995Q1–2019Q4 and 1995Q1–2024Q4. The addi￾tional period, 2020–2024, was unusually turbulent, encompassing the Covid-19 pan￾demic in 2020–2021 and the subsequent high-inflation episode beginning in 2021. Us￾ing a posterior probability threshold above 90 per cent (or below 10 per cent) to de￾fine statistical significance, we do not find any statistically significant differences in the estimated effects across the two sample periods (see Table 5). One interpretation is that, within the baseline model, including the years 2020–2024 does not materially alter the estimated effects of monetary policy. Table 5. Probability that the effect of monetary policy on GDP and inflation has increased when estimated to 2024 compared to 2019 Country GDP Inflation Australia 0.88 0.69 Canada 0.67 0.50 Chile 0.33 0.45 Czech Republic 0.54 0.61 Iceland 0.52 0.47 Norway 0.69 0.59 New Zealand 0.69 0.61 Poland 0.47 0.54 Sweden 0.44 0.58 United Kingdom 0.55 0.58 While the changes in estimated effects are not statistically significant in most cases, they may nevertheless be relevant from a monetary policy perspective. Central banks must base decisions on the best available assessment of the current transmission mechanism. From this viewpoint, the time variation in Figures 6 and 7 may be eco￾nomically meaningful even when it cannot be distinguished from earlier estimates 33 This normalisation does not capture potential changes over time in the persistence of the policy rate re￾sponse, which could affect the estimated effects, see, for example Ramey (2016).

28 with high statistical confidence. We therefore briefly comment on the patterns in the data. Figure 6. Peak effects on inflation of an increase in the policy rate estimated using sign restrictions for different samples

Note. The BVAR model is estimated for different samples, and the figure shows how the maximum effect on inflation has changed over time. The initial effect on the policy rate is normalised to one percentage point. The thin lines represent the 5th, 15th, 25th, 75th, 85th and 95th percentiles of the probability distribution of the responses. The x-axis shows the endpoint of the sample period. Ice￾land has been excluded due to difficulties in estimation with expanding samples. Overall, the estimated inflation responses are relatively stable over time. For all coun￾tries, the peak inflation response to a monetary policy tightening is negative across the alternative sample endpoints. At the same time, the estimates suggest that the magnitude of the inflation response has increased somewhat over time in a subset of countries, including New Zealand, the Czech Republic, Poland, and Sweden. The evidence for GDP is less clear-cut. In some countries, such as Australia, Norway, and New Zealand, the peak GDP effect appears to have become larger over time, alt￾hough the associated uncertainty bands imply that these changes should be inter￾preted cautiously. In both Figures 6 and 7, the pandemic years 2020–2021 introduce a visible kink in the estimated peak effects for several countries. However, extending the sample to include the post-2019 period does not appear to materially change the

29 overall magnitude of the estimated effects. Nevertheless, given the exceptional na￾ture of the pandemic period, reporting estimates both including and excluding these observations helps assess the robustness of the results to recent macroeconomic dis￾ruptions. Figure 7. Peak effects on GDP of an increase in the policy rate estimated using sign restrictions for different samples Note. The BVAR model is estimated for different samples, and the figure shows how the maximum effect on GDP has changed over time. The initial effect on the policy rate is normalised to one per￾centage point. The thin lines represent the 5th, 15th, 25th, 75th, 85th and 95th percentiles of the probability distribution of the responses. The x-axis shows the endpoint of the sample period. Ice￾land has been excluded due to difficulties in estimation with expanding samples.

30 5 A closer examination of the inflation effects In this section, we focus on the estimated effects of monetary policy on inflation, as these responses appear relatively uncertain compared with other macroeconomic variables. 34 Across the ten SOEs, inflation re￾sponses under short-run zero identification frequently exhibit a price puzzle. This raises the question of whether the pattern is driven by the choice of inflation measure or instead reflects underlying identification problems. Lastly, we decompose the inflation responses into a policy rate and an exchange rate channel. We find that the inflation responses are particularly sensitive to both the inflation measure and the identification method, with a price puzzle emerging in roughly half of the countries under short-run zero re￾strictions. Using core inflation measures that exclude volatile CPI compo￾nents or sign restrictions resolves this issue. Decomposing inflation re￾sponses shows that countries with stronger inflation responses also ex￾hibit a more pronounced exchange-rate channel. 5.1 Difficulties in estimating inflation effects using short-run zero restrictions Difficulties in identifying and estimating the inflation response in small open econo￾mies have long been recognised in the monetary policy VAR literature, and many studies have therefore employed alternative inflation measures to mitigate counter￾intuitive results. For example, in the case of Norway, most SVAR analyses rely on CPI excluding energy and tax effects (CPI-ATE) rather than headline CPI to address identifi￾cation challenges and to reduce the influence of volatile components.35 Australia pro￾vides another example, where the CPI Trimmed Mean has been used as an alternative inflation indicator in several studies.36 Similarly, for the United Kingdom, the CPI Core 34 Other variables may also be affected by measurement and sample issues. For GDP, the expanding-sample analysis in Section 4 suggests that including the pandemic and post-pandemic observations does not mate￾rially change the overall magnitude of the estimated effects, although the pandemic years introduce visible movements in some country estimates. For unemployment, cross-country comparisons should be inter￾preted with some caution because the underlying labour-market series are not fully harmonised: as shown in Table B1 in Appendix B, some countries use survey-based measures while others use registered unem￾ployment. This may affect the estimated Okun coefficients and reinforces our interpretation of the cross￾country elasticity comparisons as suggestive rather than exact structural rankings. 35 The CPI-ATE is published by Statistics Norway and refers to CPI adjusted for tax changes and excluding energy products. 36 The CPI Trimmed Mean is published by Australian Bureau of Statistics and refers to the CPI adjusted to exclude large price movements in either direction. For information about the calculation methodology, see Australian Bureau of Statistics (2025).

31 measure is commonly used. More broadly, the literature on SOEs reports considerable dispersion in estimated inflation responses.37 Based on our survey of research articles, this uncertainty appears particularly pronounced in studies for Australia, Chile, New Zealand, and the UK. The variation likely reflects not only the choice of inflation meas￾ure but also differences in identification strategies, including the use of recursive or￾derings versus sign restrictions. 38 Our own estimates reflect these challenges. The estimated effect of a monetary policy shock on inflation is highly sensitive to both the choice of inflation measure and the identification strategy employed. In this section, we therefore document how our re￾sults are affected by alternative identification assumptions and inflation measures. We then relate these findings to the broader literature on price puzzles and discuss potential mechanisms that may generate such puzzles within structural VAR frame￾works. 5.2 Inflation effects with various measures and identification strategies Figure 8 presents the country-specific inflation responses to a monetary policy shock. For some countries, inflation responses are estimated using three alternative inflation measures: headline inflation (CPIF for Sweden and CPI for the remaining small open economies), a core inflation measure excluding volatile components, and, when avail￾able, an additional measure commonly used in earlier country-specific studies (such as the Trimmed CPI in Australia). The figure shows results obtained under both identi￾fication using short-run zero restrictions and sign restrictions and also includes esti￾mates from specifications in which oil prices are included in the model.39 Broadly, our findings are consistent with existing studies that rely on similar inflation measures and identification schemes, see Table 3 in Section 2.4. 40 Several patterns emerge from the country-level evidence. First, price puzzles are common under short￾run zero restrictions and when CPI is used as the measure of inflation. In the SVAR lit￾erature, researchers distinguish between short-run price puzzles, where prices rise only temporarily following a monetary policy rate increase, and persistent or long-run price puzzles, where the price level increases more durably over several years.41 Our results reflect both types: we find short-run puzzles for countries such as Australia, Norway, and New Zealand, and more persistent puzzles for the Czech Republic, 37 See Table 6 in Section 5.3 below. 38 Previous studies have also employed other identification methods, such as FAVAR and narrative ap￾proaches. 39 In Section 5.3, we outline the rationale for including commodity prices, such as oil prices, in VAR models. 40 Table 3 in Section 2.4 compares our results with previous findings from SVAR studies in all ten small econ￾omies. For some countries, we also include estimated effects from DSGE studies, however, these are less comparable with our estimates. 41 For estimated short-run price puzzles, see, for instance, Christiano et al. (2005); for documented persis￾tent price puzzles, see, for instance, Baumeister et al. (2010) and Chen and Valcarcel (2021). Previously published results often exhibit the price puzzle in the short run, possibly because findings showing persis￾tence in the long run are more difficult to publish, given that macroeconomic theory suggests prices should eventually decline following a monetary contraction. For example, the meta-analysis by Rúsnak et al. (2013) finds evidence of publication selection against results documenting the price puzzle.

32 Iceland, and the UK.42 Importantly, for Australia, New Zealand, Norway and the UK, using a core inflation measure eliminates the puzzle, illustrating how inflation meas￾urement can materially alter the inferred effects of monetary policy. A similar pattern appears in the Czech Republic, where previous work using CPI excluding energy finds no puzzle.43 In countries where the literature has converged on a preferred price in￾dex precisely to address this issue, such as CPI-ATE in Norway or core CPI in the UK, our results closely match earlier studies. This suggests that tailored inflation measures can improve identification and avoid price puzzles in recursive SVAR models. Sweden stands out as a notable exception: none of the inflation measures we use generate a price puzzle. Moreover, the relative sizes of the impulse responses (largest for CPI, smaller for CPIF, and smallest for CPIF excluding energy) reinforce the notion that measurement choices shape the estimated inflation response.44 Second, sign-restricted identification consistently produces larger negative inflation responses. Since the negative inflation response to a policy tightening is imposed by assumption, these results are not informative about the existence of price puzzles. Nonetheless, the fact that sign-restricted inflation effects exceed those obtained with short-run zero restrictions underscores the extent to which identification assumptions influence the estimated inflation effects. Finally, we explore whether augmenting the model with commodity prices, often pro￾posed as a remedy for price puzzles, improves identification.45 Including oil prices, which track broader commodity price movements well, does not change our overall findings.46 However, it somewhat mitigates the price puzzle in Iceland (see Figure 8). This suggests that the instability of inflation responses likely stems from deeper issues related to measurement and identification rather than omitted global price dynamics. As the analysis using two-channel transmission decomposition shows, the countries for which we obtain price puzzles tend to have a weaker exchange rate channel, po￾tentially resulting in a smaller inflation response (see Section 5.4 below). 42 Bjørnland and Halvorsen (2014) estimate SVAR models for six of the countries studied here over the pe￾riod 1990-2009, using a similar identification method based on short-run zero restrictions and a sign re￾striction only for the exchange rate. They find short-run price puzzles for Australia, Canada, New Zealand, and Norway, a persistent price puzzle for the UK and no puzzle for Sweden. While we use a more recent and extended sample, our results are very similar to theirs. 43 See Borys and Horvath (2008). 44 One caveat is that, under our short-run zero restrictions, all inflation measures are restricted not to re￾spond contemporaneously to the monetary policy shock. This may understate the prevalence of price puz￾zles if some CPI components respond very quickly to policy-rate changes. Sweden is a useful example. In our specification, the CPI inflation response is rather similar to the CPIF inflation response, and no price puzzle arises. By contrast, Berggren et al. (2024) find a short-run price puzzle for Swedish CPI when allowing infla￾tion to respond contemporaneously. This difference is consistent with the interpretation that interest-rate￾related housing costs can affect CPI inflation within the quarter of the policy-rate change. The price-puzzle problem may therefore be more severe than suggested by our baseline short-run zero-restriction esti￾mates. 45 See the next section, “Potential explanations from the price puzzle literature” for a discussion of why short-run zero identification can generate price puzzles. 46 The correlation between global Brent oil price series and the IMF’s World Commodity Price Index is 0.88.

33 Figure 8. Country-specific impulse response functions with headline and core inflation measures and different identification methods Note. The number of inflation measures differ across countries and are chosen based on the previous literature (see Table B2 in Appendix B for more details). Refers to inflation measured as the annual percentage change. For Sweden, the dark blue solid line represents the CPIF, the red line the CPIF excluding energy and the light blue line the CPI. For Norway, “CPI Core” refers to CPI-ATE. For Australia, the alternative inflation measure refers to the CPI Trimmed Mean. 5.3 Potential explanations from the price puzzle literature Price puzzles appear frequently in our sample of small open economies when using short-run zero restrictions, making it essential to consider why such puzzles arise. Price puzzles have long been recognised as a central challenge in the SVAR literature,

34 and the prevailing view is that they typically reflect modelling or identification short￾comings rather than genuine features of the monetary transmission mechanism. A first explanation concerns inflation measurement, particularly the reliance on head￾line CPI. In our sample, using headline CPI consistently generates price puzzles. A gen￾eral feature of the CPI is that it includes household housing costs.47 When the policy rate increases, housing-related expenses such as mortgage interest payments or rents also rise, which in turn pushes up CPI inflation.48 Since these housing costs are in￾cluded in the CPI for all ten SOEs, this mechanism may help explain the price puzzles we observe across several countries. Evidence based on US data further suggests that this channel can account for a substantial share of the price puzzle, i.e., the counterin￾tuitive response of inflation moving in the “wrong” direction.49 Another feature of headline CPI is that it includes volatile components, most notably energy and food, so its behaviour is heavily influenced by global supply shocks. Previ￾ous studies have argued that, as a result, the CPI can distort estimated inflation re￾sponses in SVAR frameworks (Sims, 1992; Hanson, 2004). When a central bank raises interest rates in response to an oil-price-driven inflation surge, oil prices may continue rising due to global conditions, causing headline CPI to rise as well. In a VAR that does not adequately capture these contemporaneous information channels, the resulting dynamics can be misinterpreted as a price puzzle. Consistent with this interpretation, switching from headline to core inflation measures removes the puzzle in several countries, including Australia, Norway, New Zealand, and the United Kingdom. A related strand of the literature has focused on omitted variable bias, particularly the exclusion of information on commodity prices. Early contributions, such as Sims (1992), showed that price puzzles largely disappear when commodity prices are in￾cluded in the VAR. The underlying argument is that central banks may respond to sig￾nals about future inflation that are not captured by the model, causing a contraction￾ary policy shock to coincide with news of rising future prices. However, in our setting, augmenting the SVAR model with oil prices has virtually no effect on the inflation re￾sponse, and the puzzle remains. This finding aligns with earlier evidence from Hanson (2004), who finds that including commodity prices does not consistently eliminate the puzzle, and with the meta-analysis of Rusnák et al. (2013), which identifies additional sources of bias, including the failure to control for potential output and financial con￾ditions. Similarly, the seminal contribution by Giordani (2004) argues that the price puzzle often arises from omitting GDP or financial variables, and that addressing these omissions can eliminate the puzzle without relying on commodity prices. In his analy￾sis, adding oil prices has little effect, whereas controlling for GDP resolves the puzzle. 47 In fact, the Riksbank changed the inflation target variable from the CPI to the CPIF in 2017, based on the argument that changes in the policy rate have direct effects on inflation through mortgage interest rates. For a discussion of the drawbacks of the CPI and the reasons for switching to the CPIF, see Sveriges Riks￾bank (2016). 48 Again, the reason we do not obtain a price puzzle with CPI inflation in the case of Sweden is because we impose a contemporaneous zero restriction for this variable. In the absence of this restriction CPI inflation increases on impact in response to a positive policy shock. 49 Dias and Duarte (2019) estimate the effects of monetary policy on house prices and rents using a proxy￾SVAR model and a small set of U.S. macroeconomic variables for the period 1983–2017.

35 Identification assumptions themselves are also central to the prevalence of price puz￾zles. A substantial literature highlights the sensitivity of estimated inflation effects to short-run zero restrictions and lag structures. Estrella (2015) surveys a wide set of VAR specifications and shows that puzzles are strongly associated with identification based on short-run zero restrictions. Non-recursive identification schemes or models with richer information sets substantially reduce the incidence of such puzzles. Rusnák et al. (2013) similarly find that identification based on short-run restrictions increases the likelihood of observing a price puzzle. As summarised in Table 6, most previous studies on the small open economies in our sample rely on sign restrictions precisely to avoid these problems. Complementing this approach, factor-augmented VAR (FAVAR) models expand the information set. Bernanke et al. (2005) demonstrate that including a broad array of macroeconomic and financial indicators, thereby approxi￾mating the central bank’s information set, yields more standard inflation responses, suggesting that limited-information VAR models may suffer from systematic misiden￾tification of monetary policy shocks. A further explanation relates to the monetary policy rule itself. Castelnuovo and Su￾rico (2010) argue that price puzzles are more likely when the central bank responds only weakly to inflation. Using U.S. data from 1960–2006, before the adoption of in￾flation targeting, they show that contractionary policy shocks tend to generate posi￾tive price responses in periods when the policy rule’s inflation coefficient is low, re￾flecting uncertainty about the central bank’s commitment to inflation stabilisation. In our sample period, which occurs after the introduction of inflation targeting regimes, such weak responses should be less relevant in principle. Nonetheless, price puzzles persist for several countries, suggesting that the VAR may fail to capture agents’ ex￾pectations of how the central bank reacts to inflation. Thus, weak perceived inflation responsiveness cannot be ruled out as a contributing factor. Incorporating explicit measures of inflation expectations into our identification scheme may therefore offer a promising direction for future research. Table 6. Peak effects of monetary policy on inflation across selected studies using different inflation measures Country Article Model and identification Method Sample period Inflation measure Peak effect on inflation, value (quar￾ter) Australia Kim and Lim (2018) SVAR, sign restrictions 1993–2014 CPI -0.1 (3) Gibbs et al. (2018) DSGE 1991–2016 CPI -0.3 (3) Read (2022) SVAR, sign restrictions 1984–2019 Trimmed Mean CPI -1 (14) Fisher and Huh (2023) SVAR, sign restrictions 1992–2016 CPI -0.4 (1) Median VAR models

• • • -0.4 This paper SVAR, sign restrictions 1995–2024 CPI -0.3

36 Country Article Model and identification Method Sample period Inflation measure Peak effect on inflation, value (quar￾ter) Canada Champagne and Sekkel (2017) VAR, Narrative 1974–2015 CPIX -0.5 (10) Kim and Lim (2018) SVAR, sign restrictions 1992–2014 CPI -0.4 (5) Alexander and Reza (2022) SVAR, sign restrictions 1981–2018 CPI -0.5 (5) Ha and So (2023) SVAR, Chole￾sky, IV 2000–2017 CPI -1 (4) Median VAR models

• • • -0.4 This paper SVAR, sign restrictions 1995–2024 CPI -0.3 Chile Madeira and Salazar (2023) FAVAR 1996–2012 CPI -0.8 (1) Aruoba et al. (2021) BVAR, Cholesky 2001–2020 CPI -0.8 (24) Median VAR models
• • • -0.8 This paper SVAR, sign restrictions 1995–2024 CPI -0.3 Czech Re￾public Borys and Horvath (2008) SVAR, Cholesky 1998–2006 CPI excl. energy -0.3 (8) Kucharcukova et al. (2013) SVAR, sign restrictions 1998–2010 CPI -0.8 (6) Czech National Bank (2021) SVAR 2000–2021 CPI -0.4 (8) Median VAR models
• • • -0.4 This paper SVAR, sign restrictions 1995–2024 CPI -0.5 Iceland Pétursson (2023) SVAR, Cholesky 2009–2022 CPI -0.6 (10) This paper SVAR, sign restrictions 1995–2024 CPI -0.1 Norway Bjørnland (2008) SVAR, Cholesky 1993–2004 CPI-ATE -0.1 (12) Gerdrup et al. (2017) DSGE - CPI-ATE -0.1 (7) Robstad (2014) BVAR, sign restrictions 1994–2013 CPI-ATE -0.3 (6) Median VAR models
• • • -0.3 This paper SVAR, sign restrictions 1995–2024 CPI -0.4 New Zealand Bloor and Matheson (2008) BVAR, Cholesky 1990–2007 Tradable and non￾tradable CPI -0.1 (8) Culling et al. (2019) DSGE 1993–2018 CPI -0.8

37 Country Article Model and identification Method Sample period Inflation measure Peak effect on inflation, value (quar￾ter) Culling et al. (2019) SVAR, sign restrictions 1993–2018 CPI -1.2 Buckle et al. (2003) SVAR, Cholesky 1983–2002 CPI -0.8 (9) Median VAR models

• • • -0.8 This paper SVAR, sign restrictions 1995–2024 CPI -0.4 Poland Demchuk et al. (2012) SVAR, Cholesky 1997–2006 CPI -0.3 (6) Chmielewski et al. (2019) SVAR 1999–2019 HICP -0.4 (9) Greszta et al. (2023) SVAR, sign restrictions 2003–2022 CPI -0.5 (8) Median VAR models
• • • -0.4 This paper SVAR, sign restrictions 1995–2024 CPI -0.8 Sweden Laséen and Strid (2013) BVAR, Cholesky 1995–2013 CPIF -0.4 (6) Corbo and Strid (2020) MAJA (DSGE) 1995–2018 CPIF -0.2 (4) Di Casola and Iversen (2019) BVAR, Cholesky 2006–2015 CPIF -0.3 (5) Lyhagen and Shahnazarian (2023) BVAR, Cholesky 2001–2022 CPIF -0.2 (10) Berggren et al. (2024) BVAR, Cholesky 1995–2024 CPIF -0.5 (6) Almerud et al. (2024) High-frequency 2004–2024 CPIF -1 Median VAR models
• • • -0.4 This paper SVAR, sign restrictions 1995–2024 CPIF -0.9 United King￾dom Ellis et al. (2014) FAVAR, Cholesky 1992–2005 CPI -2 (7) Cloyne and Hurtgen (2016) VAR, Narrative 1975–2007 RPIXm in￾flation -1 (11) Cesa-Bianchi et al. (2020) Proxy-SVAR, High-frequency 1997–2015 CPI -0.4 (4) Brignone and Piffer (2025) DSGE - CPI -0.4 (12) Brignone and Piffer (2025) SVAR, sign restrictions
• CPI -0.6 (4) Median VAR models
• • • -0.6 This paper SVAR, sign restrictions 1995–2024 CPI -0.3

38 Note. The effects reported in the table refer to inflation measures as the annual percentage change. The policy rate effect is normalised to a 1 percentage point increase for all studies. In cases where a study reports results under multiple assumptions, we have selected the one deemed most compara￾ble to our analysis. In most cases, the effects have been manually and visually extracted from fig￾ures, which entails a potential risk of reading errors. The effects from DSGE models refer to the na￾tional central bank’s dynamic stochastic general equilibrium model. The calculation of the median for VAR models includes only BVAR and SVAR estimated employing Cholesky or sign restriction iden￾tification methods. Table 6 also reports estimates from studies using alternative identification ap￾proaches, including high-frequency, narrative and FAVAR methods, when available. These studies are not included in the country medians, which are based on VAR mod￾els using Cholesky or sign-restriction identification, but they provide useful bench￾marks for assessing whether our inflation effects are quantitatively similar to esti￾mates based on other identifying assumptions. Overall, the comparison reinforces the main message of this section: inflation responses are sensitive to both identification and measurement choices, and the range of estimates across studies is wide. 50 Taken together, the evidence shows that inflation responses to monetary policy shocks in the ten SOEs are considerably less stable, and more sensitive to modelling choices, than responses of other macroeconomic variables. Price puzzles arise fre￾quently with identification based on short-run zero restrictions, particularly when headline inflation is used, and they are not resolved by adding commodity prices. While sign-restricted approaches and carefully chosen inflation measures, such as country-specific core indices, help mitigate these issues, they do not eliminate the un￾derlying identification challenges. Overall, the results highlight that our inflation esti￾mates based on short-run zero restrictions in small open economies must be inter￾preted with caution, as even the sign of the inflation response often depends more on the modelling specification than on a robust empirical regularity. 5.4 Decomposing inflation effects into channels Above we noted that there is considerable heterogeneity in the inflation effects of monetary policy across the SOEs. In this section, we decompose the inflation effect into two distinct channels: an exchange rate channel and a policy rate (or residual) channel. The analysis is based on the impulse responses to a monetary policy shock identified using both short-run zero and sign restrictions. We apply a two-channel de￾composition, which provides a convenient method for attributing impulse responses to different transmission channels. In our setting, the response of each variable to a monetary policy shock is decomposed into the dynamic effects that originate from the impact responses of the policy rate and the real exchange rate, respectively. A formal description of the decomposition is provided in Appendix A. 51 50 See, for instance, Ellis et al. (2014) and Cloyne and Hurtgen (2016) for the United Kingdom; Madeira and Salazar (2023) for Chile; and Champagne and Sekkel (2017) for Canada. 51 For a recent comprehensive treatment of Transmission Channel Analysis (TCA) in dynamic linear models, see Wegner et al (2024). Our decomposition may be viewed as a particular case of the type of decomposi￾tions discussed by these authors.

39 A first observation is that the response of the real exchange rate is almost entirely ac￾counted for by the exchange rate channel.52 This finding supports the interpretation of this component as a genuine exchange rate channel, as it captures the part of the monetary policy transmission mechanism that operates through movements in the real exchange rate. Figure 9 presents the decomposition of the CPI inflation responses (and, in the case of Sweden, CPIF). Several patterns emerge. First, the relative importance of the policy rate and exchange rate channels differs across countries. Second, the exchange rate channel typically exerts its influence on inflation more rapidly, with effects that ap￾pear earlier and peak sooner than those associated with the policy rate channel. Third, there is a clear relationship between the size of the overall inflation response and the importance of the exchange rate channel. The Czech Republic, Iceland, Po￾land, and Sweden exhibit relatively large inflation effects, and in these countries the exchange rate channel is also comparatively strong. These are likewise the countries in which the monetary-policy-conditional exchange rate pass-through was estimated to be relatively high. By contrast, Chile and Norway display smaller inflation re￾sponses, and in both cases the exchange rate channel is considerably weaker. Finally, in the two countries with the largest inflation effects – Poland and Sweden – the policy rate channel is also relatively strong, while it is rather weak in most coun￾tries. These are likewise the countries in which the monetary-policy-conditional Phil￾lips elasticities were estimated to be relatively high. 52 The exchange rate channel accounts for nearly all of the real exchange rate response in all countries ex￾cept Australia and Sweden, where it instead explains roughly three-quarters of the total effect.

40 Figure 9. The inflation effect of monetary policy decomposed into policy rate and exchange rate channels Note. The figure shows the inflation response to a normalised policy shock decomposed into policy rate and exchange rate channel contributions, respectively. The decomposition is described in Ap￾pendix A.

41 Concluding comments This staff memo has estimated the effects of monetary policy in ten inflation targeting small open economies using a standardised SVAR framework. The cross-country ag￾gregate responses are broadly consistent with standard monetary transmission: a contractionary policy shock lowers GDP and inflation, raises unemployment, and ap￾preciates the real exchange rate. The estimated magnitudes are also broadly in line with the wider VAR literature and of a similar order of magnitude to those typically re￾ported for larger economies. At the same time, the estimated effects differ across countries, although only some of these differences are statistically distinguishable from the cross-country median. The expanding-sample analysis suggests that the esti￾mated effects on GDP and inflation are broadly stable as the sample is extended, even though point estimates increase somewhat in some countries. Finally, the inflation re￾sponses are particularly sensitive to the inflation measure and the identification strat￾egy. Price puzzles arise frequently under short-run zero restrictions but are mitigated when using core inflation measures or sign restrictions. The decomposition of the in￾flation response further suggests that countries with larger inflation effects tend to have a larger exchange-rate-related contribution to inflation. A central contribution of the memo is the analysis of monetary policy transmission elasticities. Rather than only comparing how strongly GDP, unemployment, inflation and the exchange rate respond to the policy shock, we examine the responses relative to one another through an Okun coefficient, a Phillips slope, a sacrifice ratio and ex￾change rate pass-through. This helps distinguish between countries where inflation or unemployment responses are large because GDP or the exchange rate responds strongly, and countries where they are large because the relationships between these variables are stronger. The cross-country patterns suggest that unemployment and in￾flation responses are more closely related to the corresponding transmission elastici￾ties than to the raw GDP and exchange-rate responses alone. In this sense, the elastic￾ities provide a useful summary of how policy-induced movements in GDP and the ex￾change rate are associated with labour-market and price outcomes across countries. An important topic for future research is to explain more systematically why the ef￾fects of monetary policy differ across countries. The analysis in this memo documents cross-country heterogeneity in responses and transmission elasticities, but it does not attempt to relate these differences to underlying structural characteristics. One ave￾nue would be to relate country-level responses to features such as household indebt￾edness, mortgage-market structure, labour-market flexibility and financial depth, in the spirit of cross-country studies of monetary policy transmission heterogeneity such as Deb et al. (2023), or more specific applications such as Di Casola and Grothe (2026). For small open economies, it would be particularly useful to study the determinants of the exchange-rate-related part of transmission. Import intensity, commodity expo￾sure, exchange rate pass-through and the currency in which trade is invoiced may all affect how exchange-rate movements are transmitted to consumer prices. This would connect the cross-country evidence in this memo to the broader exchange rate pass￾through literature, including work that emphasises heterogeneity across countries, in￾dustries and time periods, such as Ortega and Osbat (2020), and work on the role of

42 currency invoicing and pricing behaviour, such as Gopinath, Itskhoki and Rigobon (2010). References Acevedo, Laura and Marc Hofstetter (2023), ‘Sacrifice ratios and the income distribu￾tion: Stylized facts for OECD’, Economics Letters, vol. 232, 111329. Albertini, Julien, Güneş Kamber and Michael Kirker (2011), ‘An estimated small open economy model with fractional unemployment’, Discussion Paper Series, no. 4, Re￾serve Bank of Australia. Almerud, Jakob, Dominika Krygier, Henrik Lundvall and Mambuna Njie (2024), ’Meas￾uring Riksbank Monetary Policy: Shocks and Macroeconomic Transmission’, Sveriges Riksbank Working Paper Series, no. 445, Sveriges Riksbank. Alessandri, Piergiorgio, Òscar Jordà and Fabrizio Venditti (2025), ‘Decomposing the monetary policy multiplier’, Journal of Monetary Economics, vol. 152, 103783. Alexander, Patrick and Abeer Reza (2022), ‘Exports and the Exchange Rate: A General Equilibrium Perspective’, Staff Working Paper, no. 18, Bank of Canada. Anderl, Christina and Guglielmo Maria Caporale (2023), ‘Nonlinearities in the ex￾change rate pass-through: The Role of Inflation Expectations’, International Econom￾ics, vol. 173, pp. 86–101. Aruoba, Boragan, Andrés Fernández, Daniel Guzmán, Ernesto Pastén and Felipe Saffie (2021), ‘Monetary Policy Surprises in Chile: Measurement and Real Effects’, Central Bank of Chile. Australian Bureau of Statistics (2025), ‘Underlying Inflation Measures: Explaining the Trimmed Mean and Weighted Median’ Underlying Inflation Measures: Explaining the Trimmed Mean and Weighted Median | Australian Bureau of Statistics [Accessed 15 December 2025] Ball, Laurence M., Daniel Leigh and Prakash Loungani (2013), ‘Okun’s law: Fit at Fifty?’, Working Paper, no. 18668, National Bureau of Economic Research. Ball, Laurence and Sandeep Mazumder (2020), ‘A Phillips curve for the euro area’, Working Paper, no. 2354, European Central Bank. Barnichon, Regis and Geert Mesters (2021), ‘The Phillips multiplier’, Journal of Mone￾tary Economics, vol. 117, pp. 689–705. Baumeister, Christine, Philip Liu and Haroon Mumtaz (2010), ‘Changes in the trans￾mission of monetary policy: evidence from a time-varying factor augmented VAR’, Working Paper, no. 401, Bank of England.

43 Beckers, Benjamin (2020), ‘Credit Spreads, Monetary Policy and the Price Puzzle’, Re￾search Discussion Paper, no. 1, Reserve Bank of Australia. Berkelmans, Leon (2005), ‘Credit and Monetary Policy: An Australian SVAR’, Research Discussion Paper, no. 6, Reserve Bank of Australia. Bernanke, Ben S., Jean Boivin and Piotr Eliasz (2005), ’Measuring the Effects of Mone￾tary Policy: A Factor-Augmented Autoregressive (FAVAR) Approach’, Quarterly Journal of Economics, vol. 120, no. 1, pp. 287–422. Berggren, Erik, Stefania Mammos and Ingvar Strid (2024), ’Penningpolitikens effekter i Sverige under inflationsmålsperioden: skattningar med strukturella VAR-modeller’, Staff Memo, Sveriges Riksbank. Bhuiyan, Rokon (2012), ’Monetary transmission mechanisms in a small open econ￾omy: A Bayesian structural VAR approach’, The Canadian Journal of Economics, vol. 45, no. 3, pp. 1037–1061. Bjørnland, Hilde C., Leif Brubakk and Nicolò Maffei-Faccioli (2024), ‘Piecing the Puzzle: Real Exchange Rates and Long-Run Fundamentals’, Working Paper, no. 21, Norges Bank. Bjørnland, Hilde C. (2008), ‘Monetary Policy and Exchange Rate Interactions in a Small Open Economy’, The Scandinavian Journal of Economics, vol. 110, no. 1, pp. 197–221. Bjørnland, Hilde C. and Jørn Halvorsen (2014), ‘How does monetary policy respond to exchange rate movements? New international evidence’, Oxford Bulletin of Economics and Statistics, vol 76, 2, pp. 208–232. Bloor, Chris and Troy Matheson (2008), ‘Analysing shock transmission in a data-rich environment: A large BVAR for New Zealand’, Discussion Paper Series, no. 9, Reserve Bank of New Zealand. Borys, Magdalena M. and Roman Horváth (2008), ’The Effects of Monetary Policy in the Czech Republic: An Empirical Study’, Working Paper Series, no. 4, Czech National Bank. Brignone, Davide and Michela Piffer (2025), ‘A structural VAR model for the UK econ￾omy’, Macro Technical Paper, no. 3, Bank of England. Brischetto, Andrea and Graham Voss (1999), ‘A Structural Vector Autoregression Model of Monetary Policy in Australia’, Research Discussion Paper, no. 11, Reserve Bank of Australia. Buckle, Robert A., Kim Kunghong and Nathan McLellan (2003), ‘The impact of mone￾tary policy on New Zealand business cycles and inflation variability’, Working Paper Series, no. 523214v2, New Zealand Treasury. Castelnuovo, Efrem and Paolo Surico (2010), ‘Monetary Policy, Inflation Expectations and the Price Puzzle’, Economic Journal, vol. 120, no. 549, pp. 1262–1283.

44 Cesa-Bianchi, Ambrogio, Gregory Thwaites and Alejandro Vicondoa (2020), ‘Monetary policy transmission in the United Kingdom: A high frequency identification approach’, European Economic Review, vol. 123. Champagne, Julien and Rodrigo Sekkel (2017), ‘Changes in Monetary Regimes and the Identification of Monetary Policy Shocks: Narrative Evidence from Canada’, Staff Working Paper, no. 39, Bank of Canada. Chen, Zhengyang and Victor J. Valcarcel (2021), ‘Monetary transmission in money markets: The not-so-elusive missing piece of the puzzle’, Journal of Economic Dynam￾ics and Control, vol. 131. Chmielewski, Tomasz, Andrzej Kociecki, Tomasz Lyziak, Jan Przystupa, Ewa Stani￾slawska, Malgorzata Waleryck and Ewa Wrobel (2019), ‘Monetary policy transmission mechanism in Poland. What do we know in 2019?’, Working Paper, no. 329, National Bank of Poland. Christiano, Lawrence J., Martin Eichenbaum and Charles L. Evans (2005), ‘Nominal Ri￾gidities and the Dynamic Effects of a Shock to Monetary Policy’, Journal of Political Economy, vol. 113, no. 1, pp. 1–45. Cloyne, James and Patrick Hurtgen (2016), ‘The Macroeconomic Effects of Monetary Policy: A New Measure for the United Kingdom’, American Economic Journal, vol. 8, no. 4, pp. 75–102. Corbo, Vesna and Paola Di Casola (2022), ‘Drivers of consumer prices and exchange rates in small open economies’, Journal of International Money and Finance, vol. 122, no. 102553. Corbo, Vesna and Ingvar Strid (2020), ’MAJA: A two-region DSGE model for Sweden and its main trading partners’, Working Paper, no. 391, Sveriges Riksbank. Culling, Jamie, Punnoose Jacob, Adam Richardson, Evelyn Truong and Tugrul Vehbi (2019), ‘Have the effects of monetary policy on inflation and economic activity in New Zealand changed over time?’, Analytical Note Series, no. 10, Reserve Bank of New Zealand. Czech National Bank (2021). ‘Monetary Policy Transmission in the Czech Economy’, Monetary Policy Report. Daly, Marc C., John G. Fernald, Óscar Jordá and Fernanda Nechio (2014), ‘Labor Mar￾kets in the Global Financial Crisis: The Good, the Bad and the Ugly’, Working Paper Se￾ries, no. 11, Federal Reserve Bank of San Francisco. Deb, Pragyan, Julia Estefania-Flores, Melih Firat, Davide Furceri, and Siddharth Kothari (2023), ‘Monetary Policy Transmission Heterogeneity: Cross- Country Evidence’, Working Paper, no. 23/204, International Monetary Fund.

45 Demchuk, Oksana, Tomasz Lyziak, Jan Przystupa, Anna Sznajderska and Ewa Wrobel (2012), ‘Monetary policy transmission in Poland. What do we know in 2011?’, Work￾ing Paper, no. 116, National Bank of Poland. Dias, Daniel A. and João D. Duarte (2019), ‘Monetary policy, housing rents, and infla￾tion dynamics’, Journal of Applied Econometrics, vol. 34, no. 5, pp. 673–687. Di Casola, Paola and Magdalena Grothe (2026), ‘Housing Wealth and Monetary Policy Transmission: Cross-Country Evidence’, ECB Working Paper, no. 2026/3204, European Central Bank. Di Casola, Paola and Jens Iversen (2019), ‘Monetary policy with high household debt and low interest rates’, Staff Memo, Sveriges Riksbank. Enzinger, Matthias, Sebastian Gechert, Philipp Heimberger, Franz Prante and Daniel F. Romero (2025), ‘The overstated effects of conventional monetary policy on output and prices’, 14R Discussion Paper Series, no. 264, Institute for Replication (14R). Ellis, Colin, Haroon Mumtaz and Pavel Zabczyk (2014), ‘What Lies Beneath? A Time￾Varying FAVAR Model for the UK Transmission Mechanism’, The Economic Journal, vol. 124, pp. 668–699. Estrella, Arturo (2015), ‘The Price Puzzle and VAR Identification’, Macroeconomic Dy￾namics, vol. 19, no. 7, pp. 1543–1567. Fisher, Lance A. and Hyeon-seung Huh (2023, ‘Systematic monetary policy in a SVAR for Australia’, Economic Modelling, vol. 128. Foroni, Claudia and Francesco Furlanetto (2022), ‘Explaining deviations from Okun’s law’, Working Paper, no. 2699, European Central Bank. Gerdrup, Karsten R., Erling Kravik Motzfeldt, Kenneth Paulsen Sæterhagen and Ørjan Robstad (2017), ‘Documentation of NEMO – Norges Bank’s Core Model for Monetary Policy Analysis and Forecasting’, Staff Memo, no. 8, Norges Bank. Gibbs, Christopher G., Jonathan Hambur and Gabriela Nodari (2018), ‘DSGE Reno: Adding a Housing Block to a Small Open Economy Model’, Research Discussion Paper, no. 4, Reserve Bank of Australia. Giordani, Paolo (2004), ‘An alternative explanation of the price puzzle’, Journal of Monetary Economics, vol. 51, no. 1, pp. 1271–1296. Gopinath, Gita, Oleg Itskhoki and Roberto Rigobon (2010), ‘Currency Choice and Ex￾change Rate Pass-Through’, American Economic Review, vol. 100, no. 1, pp. 304–336. Greszta, Michal, Marcin Humanicki, Mariusz Kapuscinski, Tomasz Kleszcz, Andrzej Ko￾ciecki, Jacek Kotlowski, Michal Ledochwski, Michal Lesyk, Tomasz Lyziak, Mateusz Pipien, Piotr Popowski, Ewa Stansislawska, Karol Szafranek, Grzegorz Szafranksi, Dorota Scibisz, Grzegorz Wesolowski and Ewa Wrobel (2023), ‘Monetary policy

46 transmission mechanism in Poland. What do we know in 2023?’, Working Paper, no. 365, National Bank of Poland. Ha, Jongrim and Inhwan So (2023), ‘Which Monetary Shocks Matter in Small Open Economies? Evidence from Canada’, International Journal of Central Banking. Hanson, Michael S. (2004), ‘The ‘price puzzle’ reconsidered’, Journal of Monetary Eco￾nomics, vol. 51, no.1, pp. 1385–1413. Haug, Alfred A., Tomasz Jędrzejowicz and Anna Sznajderska (2019), ‘Monetary and fis￾cal policy transmission in Poland’, Economic Modelling, vol. 79, pp. 15–27. Hopkins, Elizabeth, Jesper Lindé and Ulf Söderström (2009), ’The monetary transmis￾sion mechanism’, Economic Review, no. 2, Sveriges Riksbank. Jonsson, Magnus and Anders Vredin (2025), ’30 years of inflation targeting: from sim￾ple to complex’, Economic Review, no. 1, Sveriges Riksbank. Jašová, Martina, Richhild Moessner and Elód Takáts (2016), ‘Exchange rate pass￾through: What has changed since the crisis?’, Working Paper, no. 583, Bank for International Settlements. Jääskelä, Jarkko and David Jennings (2010), ’Monetary Policy and the Exchange Rate: Evaluation of VAR Models’, Research Discussion Paper, no. 7, Reserve Bank of Aus￾tralia. Kim, Soyoung and Kuntae Lim (2018), ‘Effects of monetary policy shocks on exchange rates in small open Economies’, Journal of Macroeconomics, vol. 56, pp. 324–339. Kiley, Michael T. and Frederic S. Mishkin (2025), ‘The Evolution of Inflation Targeting from the 1990s to 2020s: Developments and New Challenges’, Finance and Economics Discussion Series, no. 25. Washington: Board of Governors of the Federal Reserve Sys￾tem. Kucharcukova, Oxana Babecká, Michal Franta, Dana Hájková, Petr Král, Ivana Ku￾bicová, Anca Podpiera and Branislav Saxa (2013), ‘What We Know About Monetary Policy Transmission in the Czech Republic: Collection of Empirical Results’, Research and policy notes, no. 1, Czech National Bank. Laséen, Stefan and Ingvar Strid (2013), ’Debt Dynamics and Monetary Policy: A Note’, Working Paper, no. 283, Sveriges Riksbank. Lyhagen, Johan and Hovick Shahnazarian (2023), ’Samspelet mellan finans- och pen￾ningpolitik i Sverige’, Bilaga 5 till Långtidsutredningen 2023. Demchuk, Oksana, Tomasz Lyziak, Jan Przystupa, Anna Sznajderska and Ewa Wróbel (2011), ’Monetary policy transmission mechanism in Poland. What do we know in 2011?’, Working Paper, no. 116, National Bank of Poland.

47 Mazumder, Sandeep (2014), ’Determinants of the sacrifice ratio: Evidence from OECD and non-OECD countries’, Economic Modelling, vol. 40, pp. 117–135. Madeira, Carlos and Leonardo Salazar (2023), ‘The Impact of Monetary Policy on La￾bor Market with Heterogeneous Workers: The Case of Chile’, Latin American Journal of Central Banking, vol. 2, no. 2, 100092. Ortega, Eva and Chiara Osbat (2020), ‘Exchange Rate Pass-Through in the Euro Area and EU Countries’, Documentos Ocasionales, no. 2016, Banco de España. Pétursson, Thórarinn G. (2023), ‘Monetary Policy Transmission in Iceland: Evidence from a Structural VAR Model’, Working Paper, no. 94, Central Bank of Iceland. Raghavan, Mala, George Athanasopoulos and Param Silvapulle (2016), ‘Canadian monetary policy analysis using a structural VARMA model’, The Canadian Journal of Economics, vol. 49, no. 1, pp. 347–373. Ramey, V.A (2016), ‘Chapter 2 – Macroeconomic Shocks and Their Propagation’, in Handbook of Macroeconomics (vol. 2), edited by John B. Taylor and Harald Uhlig, Else￾vier. Read, Matthew (2022), ‘Estimating the Effects of Monetary Policy in Australia Using Sign-restricted Structural Vector Autoregressions’, Research Discussion Paper, no. 9, Reserve Bank of Australia. Robstad, Ørjan (2014), ‘House prices, credit and the effect of monetary policy in Nor￾way: Evidence from structural VAR Models’, Working Paper, no. 5, Norges Bank. Rogoff, Kenneth (1985), ‘The optimal degree of commitment to a monetary target’, Quarterly Journal of Economics, vol. 100, pp. 1169–1190. Rusnák, Marek, Tomas Havranek and Roman Horvath (2013), ‘How to solve the Price Puzzle? A Meta-Analysis’, Journal of Money, Credit and Banking, vol. 45, no. 1, pp. 37– 70. Sims, Christopher A. (1980), ‘Macroeconomics and Reality’, Econometrica, vol. 48, no. 1, pp. 1–48. Sims, Christopher A. (1986), ‘Are Forecasting Models Usable for Policy Analysis?’, Fed￾eral Reserve Bank of Minneapolis Quarterly Review, Winter 1986, pp. 2-16. Sims, Christopher A. (1992), ‘Interpreting the macroeconomic time series facts: the ef￾fects of monetary policy’, European Economic Review, vol. 36, no. 5, pp. 975–1000. Svensson, Lars E.O (2007), ’Inflation Targeting’, in The New Palgrave Dictionary of Eco￾nomics (2nd edition), edited by Larry Blum and Steven Durlauf, Princeton University Press. Sveriges Riksbank (2016), ‘The Riksbank’s inflation target – target variable and inter￾val’, September 2016, Riksbank Studies.

48 Tetlow, Robert J. (2022), ‘How Large is the Output Cost of Disinflation?’, Finance and Economics Discussion Series, no. 2022-079, Board of Governors of the Federal Re￾serve System. Villani, Mattias and Anders Warne (2003), ‘Monetary policy analysis in a small open economy using Bayesian cointegrated structural VARs, Working Paper Series, no. 296, European Central Bank. Wegner, Enrico, Lenard Lieb, Stephan Smeekes and Ines Wilms (2024), ‘Transmission Channel Analysis in Dynamic Models’, arXiv preprint arXiv:2405.18987. Ziegenbein, Alexander (2021), ‘Macroeconomic shocks and Okun’s Law’, Economic Letters, vol. 202.

APPENDIX 49 APPENDIX APPENDIX A describes the block exogenous VAR model and short-run zero identification approach. A description of the method for decompos￾ing the inflation effect of a monetary policy shock into two channels – a policy rate channel and an exchange rate channel – is also provided. APPENDIX B describes the data and transformations used to construct the macroeconomic variables used in the VAR model, as summarised in Table B1. Table B2 presents an overview of previous related country-spe￾cific studies, and the different inflation and policy-rate measures they employ. Table B3 presents the trade weights used to construct the coun￾try-specific foreign variables. APPENDIX C provides an exact description of the impulse response statis￾tics (e.g. median peak effects for individual countries and aggregated across countries) reported in the paper. APPENDIX D presents additional figures from the analysis. Figure D1 dis￾plays country-specific impulse response functions under sign restrictions, while Figure D2 shows the corresponding responses under short-run zero restrictions. Figure D3 reports the peak effects of monetary policy for the ten SOEs using short-run zero restrictions, and Figure D4 illustrates the cumulative exchange rate pass-through in the ten SOEs. APPENDIX E describes the regression estimates of the Okun coefficient, Phillips slope, and exchange rate pass-through that are used to derive the monetary policy transmission elasticities in Section 3. Table E1 pre￾sents comparisons of MP-conditional elasticities using different calcula￾tion methods. APPENDIX A – VAR, short-run zero identification, moving average representation and two-channel decomposition The reduced form VAR(K) model is 𝑦𝑡 = 𝑐 + 𝐴1𝑦𝑡−1 + 𝐴2𝑦𝑡−2 + ⋯ + 𝐴𝐾𝑦𝑡−𝐾 + 𝑢𝑡 with 𝐸(𝑢𝑡 ) = 0, 𝐸(𝑢𝑡𝑢𝑡 ′ ) = Σ𝑢, 𝐸(𝑢𝑡𝑢𝑡−𝑠 ′ ) = 0, 𝑠 ≠ 0

APPENDIX 50 where 𝑦𝑡 is a 𝑃 × 1 vector of variables, 𝐴𝑘 𝑘 = 1, . . ,𝐾, are 𝑃 × 𝑃 coefficient matrices and 𝑢𝑡 is the 𝑃 × 1 vector of reduced form innovations. Short-run zero identification introduces orthonormal structural shocks 𝜀𝑡 such that 𝑢𝑡 = 𝐵𝜀𝑡 , 𝐸(𝜀𝑡𝜀𝑡 ′ ) = 𝐼, which implies 𝛴𝑢 = 𝐸(𝑢𝑡𝑢𝑡 ′ ) = 𝐵𝐵 ′ . Under short-run zero identification, 𝐵 is taken to be lower triangular, with the order￾ing of variables in 𝑦𝑡 determining the implied contemporaneous restrictions. We assume that the foreign variables are block exogenous. With three foreign varia￾bles placed first in 𝑦𝑡 the upper-right block of 𝐴𝑘 of dimension 3 × (𝑃 − 3) is zero (foreign variables do not depend on lagged domestic variables). We further assume that the real exchange rate may respond contemporaneously to the monetary policy shock while the other variables respond with a lag. Accordingly, we order the variables so that the real exchange rate is placed last in the vector 𝑦𝑡 , (position 𝑃), and the policy rate is placed in position 𝑃 − 1. In our baseline model we have 𝑃 = 8 variables. Let the monetary policy shock be 𝜀𝑃−1,𝑡 (shock index 𝑃 − 1). Under the recursive structure, the impact (horizon ℎ = 0) responses are read from 𝑢𝑡 = 𝐵𝜀𝑡 . Holding all other shocks at zero, a one-unit policy shock implies ∆𝑦𝑃−1,𝑡 = 𝑢𝑃−1,𝑡 = 𝑏𝑃−1,𝑃−1𝜀𝑃−1,𝑡 and ∆𝑦𝑃,𝑡 = 𝑢𝑃,𝑡 = 𝑏𝑃,𝑃−1𝜀𝑃−1,𝑡 while 𝑢𝑗,𝑡 = 0 for 𝑗 < 𝑃 − 1. Hence, on impact the monetary policy shock enters the system only through the reduced-form innovations of the policy-rate equation (𝑢𝑃−1,𝑡 ) and the exchange-rate equation (𝑢𝑃,𝑡 ). Other domestic variables (e.g. GDP, unemployment, inflation) do not respond contemporaneously to 𝜀𝑃−1,𝑡 . Because the foreign block is block-exogenous, the foreign variables are not affected by domestic shocks in the dynamic system. Now, the responses to the monetary policy shock are fully determined by its contem￾poraneous impact on the reduced-form innovations in the policy-rate and exchange￾rate equations. Under the recursive identification with the policy rate ordered in posi￾tion 𝑃 − 1 and the real exchange rate in position 𝑃, the monetary policy shock loads contemporaneously only on 𝑢𝑃−1,𝑡 and 𝑢𝑃,𝑡 . This motivates a decomposition of the impulse responses to the monetary policy rate shock into two channels; an exchange rate channel (the component transmitted through 𝑢𝑃,𝑡 ) and a policy rate (or residual) channel (the component transmitted through 𝑢𝑃−1,𝑡 ). We may note that for most

APPENDIX 51 countries in our study the movement in the real exchange rate following a monetary policy shock is almost entirely accounted for by the exchange rate channel. This, we believe, validates the decomposition. Next, we show how the decomposition is obtained using the moving average (MA) representation which is provided by 𝑦𝑡 − μ = ∑Γℎ ∞ ℎ=0 𝑢𝑡−ℎ = ∑Γℎ ∞ ℎ=0 𝐵𝜀𝑡−ℎ = ∑𝛩ℎ ∞ ℎ=0 𝜀𝑡−ℎ where 𝛤ℎ = ∑ 𝐴𝑗𝛤ℎ−𝑗 𝑚𝑖𝑛(𝐾,ℎ) 𝑗=1 , ℎ ≥ 1 and 𝛤0 = 𝐼 and 𝛤ℎ = 0 for ℎ < 0. In our case the impulse response to the policy shock (in position 𝑃 − 1) for domestic variable 𝑖 at horizon ℎ may be written as the sum of the two channels 𝛩𝑖,𝑃−1,ℎ = Γ𝑖,𝑃−1,ℎb𝑃−1,𝑃−1 + Γ𝑖,𝑃,ℎb𝑃,𝑃−1 The first term captures the part of the response that originates from the contempora￾neous impact of the policy shock on the policy-rate innovation 𝑢𝑃−1,𝑡 propagated through the VAR dynamics. The second term captures the part that originates from the contemporaneous impact of the same policy shock on the exchange-rate innova￾tion 𝑢𝑃−1,𝑡 , also propagated through the VAR dynamics. We refer to these as the pol￾icy-rate (residual) channel and the exchange-rate channel, respectively. APPENDIX B – Data and transformations The variables used in the model, together with their respective sources and transfor￾mations, are presented in Table B1 below. The measures used for GDP, inflation, the real effective exchange rate, the unemployment rate, and the foreign variables are largely consistent across all ten SOEs. Using the same measures for these variables en￾sures internal consistency and facilitates cross-country comparisons of the effects of monetary policy on key macroeconomic indicators. In our baseline VAR model, inflation is measured using the CPI for all countries except Sweden, where the CPIF is applied. In the in-depth analysis of inflation responses, we additionally employ core inflation measures and other indicators frequently used in the literature to better understand the underlying drivers of price dynamics. In some specifications, we include oil prices.53 Table B2 provides an overview of the inflation measures used in the previous literature. 53 See Section 5.3 for a discussion on the rationale for including oil prices in structural VAR models.

APPENDIX 52 However, the choice of monetary policy interest rates differs across countries, reflect￾ing conventions in the SVAR literature.54 Previous research shows that while some countries rely on the central bank’s official policy rate, others use short-term market rates as proxies.55 In the Czech Republic, Norway, New Zealand, and Poland, the short-term domestic effective nominal interest rate is employed, whereas in the re￾maining countries the central bank’s policy rate remains the standard measure. This choice is due to the lack of policy rate data covering the entire sample period for the Czech Republic, New Zealand and Poland, while for Norway, it follows established lit￾erature conventions.56 To ensure comparability, our model follows a standardised approach. The foreign vari￾ables (GDP, CPI, and the policy rate) are largely constructed in the same way for all countries. For most countries, they are calculated as trade-weighted aggregates of the United States and the euro area, using time-varying, country-specific trade weight data from the national central banks. The historical average country trade weights are reported in Table B3, offering an indication of the relative trade shares between the US and the euro area for each country. For the remaining countries, we use either the United States or the euro area as the foreign aggregate. For Australia and Chile, the United States is used, while for Poland the euro area is chosen. These selections re￾flect the predominant trading partners of each country. As shown in Table B2, the use of foreign variables in previous VAR models varies across countries. 54 These studies were selected based on their relevance to our VAR framework, identification strategy, and connection to central bank research. 55 See Table B2 below for an overview of previous studies and the various policy rate measures they em￾ploy. 56 The correlation between policy rate and short-term domestic effective nominal interest rate is above 0.98 for the Czech Republic, Norway, New Zealand and Poland.

APPENDIX 53 Table B1. Data and transformations Key macroeconomic variables Sources Transformations Domestic variables Inflation measured by the Con￾sumer Price Index (CPI). For Sweden, the CPIF was ap￾plied. Australian Bureau of Statistics, Chil￾ean National Statistics Institute, Czech Statistical Office, Latin Macro Watch, Polish Central Statistical Of￾fice, Statistics Canada, Statistics Iceland, Statistics New Zealand, Statistics Norway, Statistics Sweden and U.K. Office for National Statis￾tics. First difference of the natural logarithm. Inflation measures for each country were seasonally adjusted by the Riksbank using the X-11 ARIMA method developed by the U.S. Census Bureau. Inflation data were converted into quarterly values. For Chile, the series from the Chilean Na￾tional Statistics Institute were extended backwards from 2009 to 1995, using data from Latin Macro Watch. Gross Domestic Product in con￾stant prices, quarterly data. Seasonally adjusted data for all countries, and for Sweden they are additionally calendar-ad￾justed. For Norway, the measure used is mainland GDP, which ex￾cludes the oil-based offshore sector. This measure is used in comparable studies, as it is con￾sidered to better reflect the country’s export-oriented econ￾omy (see, for example, Rob￾stad, 2014; Bjørnland, 2008; Bjørnland et al., 2024). Australian Bureau of Statistics, Cen￾tral Bank of Chile, Eurostat, OECD, Statistics Canada, Statistics Iceland, Statistics New Zealand, Statistics Norway, Statistics Sweden, Sveriges Riksbank and U.K. Office for Na￾tional Statistics. Natural logarithm. For Chile, the series from the Central Bank of Chile were extended backwards from 1996 to 1995 using data from the OECD. In this pro￾cess, the GDP growth rate was calculated based on the OECD series, which was then employed to reconstruct historical values back to the first quarter of 1995. Unemployment rate in per cent at the quarterly level. Sea￾sonally adjusted data. While the age groups are broadly similar across coun￾tries, some differences exist; in Canada and Chile, 15 years and over; in the United Kingdom, 16 years and over; and in Australia 18-74 years. In Norway, New Zealand, the Czech Republic, Poland, and Iceland, the unem￾ployment rate is instead based on the total number of regis￾tered unemployed citizens. Australian Bureau of Statistics, Czech Ministry of Labor and Social Affairs, Icelandic Directorate of La￾bor, Norwegian Labour and Wel￾fare Administration, OECD, Polish Ministry of Family, Labour and So￾cial Policy, Statistics Canada, Statis￾tics New Zealand, Statistics Swe￾den, Sveriges Riksbank and U.K. Of￾fice for National Statistics. Level. For the Czech Republic, Norway, and Poland, the unemployment data were seasonally ad￾justed by the Riksbank using the X-11 ARIMA method developed by the U.S. Census Bu￾reau. For the remaining countries, seasonally adjusted data were obtained from national statistical agencies. Quarterly data were available for New Zea￾land and Sweden, while for the remaining countries (Australia, Canada, Chile, the Czech Republic, Iceland, Norway, Poland, and the United Kingdom), monthly data were con￾verted into quarterly series. Real effective exchange rate For each country, we use the broad index for the real effec￾tive exchange rate developed by the Bank for International Settlements (BIS). The BIS ef￾fective exchange rate indices cover 64 economies and are calculated as geometric trade￾weighted averages of bilateral Bank for International Settlements. Natural logarithm. The original series are at the monthly fre￾quency and were converted by the Riksbank into quarterly values.

APPENDIX 54 Table B1. Data and transformations exchange rates adjusted for rel￾ative consumer prices. Policy rate in per cent and quarterly averages. AU: Cash Rate CA: Overnight Target Rate CL: Policy rate CZ: 3-month PRIBOR IC: Key Interest Rate NO: 3-month NIBOR NZ: 90-day Bank Bill Rate PL: 1-month PRIBOR SE: Policy rate UK: Bank Rate Bank of England, Central Bank of Iceland, National Bank of the Czech Republic, Norske Finansiella Refer￾anser AS, Polish GPW Benchmark S.A., Reserve Bank of Australia, Re￾serve Bank of Chile, Reserve Bank of New Zealand, Statistics Canada, and Sveriges Riksbank. Level. For Sweden, the Riksbank publishes policy rate data at the quarterly level. For all other countries, daily rates were converted into quarterly averages by the Riksbank. Foreign variables United States Gross Domestic Product in constant prices (USD) at the quarterly level. Seasonally adjusted data. U.S. Bureau of Economic Analysis. Natural logarithm. United States Consumer Price Index (CPI) at the quarterly level. Seasonally adjusted data. U.S. Bureau of Labour Statistics. First difference of the natural logarithm. Monthly inflation data were converted into quarterly values. United States Policy Rate in per cent and quarterly aver￾ages. Federal Reserve. Level. Daily rates were converted into quarterly av￾erages. Euro Area Gross Domestic Product in constant prices (EUR) at the quarterly level. Eurostat. Natural logarithm. Euro Area Harmonised Indices of Consumer Prices (HICP) at the quarterly level. Seasonally adjusted data. ECB and Eurostat. First difference of the natural logarithm. Since the seasonally adjusted HICP series from the ECB is only available from January 1997 onwards, the series is extended back to January 1995, using seasonally adjusted his￾torical HICP data from Eurostat. The two se￾ries are linked based on quarterly percentage changes. The historical HICP data from Euro￾stat are seasonally adjusted using the X-11 ARIMA method developed by the U.S. Census Bureau. Euro Overnight Index Average (EONIA) in per cent and quar￾terly averages. EONIA is the average overnight reference rate for which Euro￾pean banks lend to one another in euros. The Riksbank. Level. Oil prices measured as spot price of Brent crude oil in USD. Intercontinental Exchange and the Riksbank. First difference of the natural logarithm. Daily rates converted into quarterly values.

APPENDIX 55 Table B2. Variables used in the previous VAR literature Country Domestic inflation Domestic policy rate Foreign variables Australia CPI (Gibbs et al., 2018; Brischetto and Voss, 1999; Berkelmans, 2005; Fisher and Huh, 2023) Underlying CPI (Beckers, 2020) Trimmed Mean CPI (Read, 2022, Jääskelä and Jennings, 2010) RBA Cash rate (Gibbs et al., 2018; Berkelmans, 2005; Jä￾äskelä and Jennings, 2010; Fisher and Huh, 2023; Read, 2022; Beckers, 2020). Foreign GDP: GDP of Australia’s largest major trading partners weighted by export share (Gibbs et al., 2018). US GDP (Jääskelä and Jennings, 2010; Read, 2022). Foreign inflation: Inflation rate of Australia’s major trading partners weighted by export share (Gibbs et al., 2018), US infla￾tion (Jääskelä and Jennings, 2010; Fisher and Huh, 2023) Foreign policy rate: Policy rate of the US, Japan and euro area (Gibbs et al., 2018), US policy rate (Jääskelä and Jennings, 2010; Fisher and Huh, 2023; Read, 2022; Beckers, 2020). Canada CPI (Bhuiyan, 2012, Raghavan et al., 2016; Alex￾ander and Reza, 2022; Ha and So, 2023; Kim and Lim, 2018) CPIX (Champagne and Sekkel, 2017) Overnight Target Rate (Bhui￾yan, 2012; Raghavan et al., 2016; Alexander and Reza, 2022; Ha and So, 2023) Foreign GDP: US Industrial production (Bhuiyan, 2012; Ha and So, 2023; Kim and Lim, 2018), US GDP (Alexander and Reza, 2022) Foreign inflation: US inflation (Bhuiyan, 2012; Bjørnland et al., 2024; Ha and So, 2023; Kim and Lim, 2018) Foreign policy rate: US policy rate (Bhuiyan, 2012; Raghavan et al., 2016; Ha and So, 2023; Kim and Lim, 2018) Chile CPI (Madeira and Salazar, 2023, Auroba et al., 2021) 1-year real interest rate (Ma￾deira and Salazar, 2023) Policy rate (Aruoba et al., 2021) Previous models do not include for￾eign variables. Czech Republic CPI (Kucharcukova et al., 2013; Czech National Bank, 2021) CPI excluding regulated prices (Borys and Horvath, 2008) 3-month PRIBOR (Borys and Horvath, 2008; Kucharcukova et al., 2013, Czech National Bank, 2021) Foreign policy rate: 1y EURIBOR (Bo￾rys and Horvath, 2008) Iceland CPI (Pétursson, 2023) Key Interest Rate (Pétursson, 2023) Foreign policy rate: Trade-weighted average foreign monetary policy rate (Pétursson, 2023) Norway CPI-ATE (Bjørnland, 2008; Gerdrup, 2017; Robstad, 2014) CPI (Bjørnland et al., 2024) 3-month NIBOR (Bjørnland, 2008; Bjørnland et al., 2024; Bjørnland et al. 2024; Rob￾stad, 2014) Nominal sight deposit rate (Gerdrup, 2017) Foreign GDP: GDP in 25 of Norway’s most important trading partners (Gerdrup, 2017) Foreign inflation: Trade-weighted CPI for 25 of Norway’s trading part￾ners (Gerdrup, 2017; Robstad, 2014) Foreign policy rate: Trade-weighted foreign interest rate based on Nor￾way’s four largest trading partners

APPENDIX 56 Table B2. Variables used in the previous VAR literature (Bjørnland, 2008; Bjørnland et al., 2024), 3-month nominal money mar￾ket interest rates for trading part￾ners (Gerdrup, 2017) New Zealand CPI (Albertini et al., 2011; Buckle et al., 2003; Culling et al., 2019) Tradable and non-tradable CPI (Bloor and Matheson, 2008) 90-day Bank Bill Rate (Alber￾tini et al., 2011; Bloor and Matheson, 2008; Culling et al., 2019) Nominal interest rate (Buckle et al., 2003) Foreign GDP: Trade-weighted meas￾ure of the GDP of New Zealand’s 16 largest trading partners (Albertini et al., 2011) World GDP (Bloor and Matheson, 2008), Trade-weighted world industrial production (Buckle et al., 2003) Foreign inflation: Trade-weighted measure of the CPI inflation of New Zealand’s 16 largest trading-partners (Albertini et al., 2011), World CPI (Bloor and Matheson, 2008) Foreign policy rate: 80-20 weighted measure of the US 90-day bank bill yield and the Aus￾tralian 90-day bank bill yield (Alber￾tini et al., 2011), World 90-day rates (Bloor and Matheson, 2008), weighted average of Australian, US, UK, Japan and German 90-day rates (Buckle et al., 2003). Poland CPI (Greszta et al., 2023; Haug et al., 2019) HICP (Demchuk et al., 2011) 3-month WIBOR (Greszta et al., 2023; Haug et al., 2019) Inter-bank short-term inter￾est rate (Demchuk et al., 2011) Foreign inflation: Euro-area HICP (Haug et al., 2019) Foreign policy rate: 1-month EURI￾BOR (Demchuk et al., 2011) Sweden CPIF (Lyhagen and Shahnazarian, 2023; Di Casola and Iversen, 2019; Laséen and Strid, 2013; Hop￾kins et al., 2009; Villani and Warne, 2003) Policy rate (Lyhagen and Shahnazarian, 2023; Di Casola and Iversen, 2019; Laséen and Strid, 2013; Hop￾kins et al., 2009; Villani and Warne, 2003) Foreign GDP: Trade-weighted meas￾ure of foreign GDP (Laséen and Strid, 2013; Di Casola and Iversen, 2019; Hopkins et al., 2009; Villani and Warne, 2003) Foreign inflation: Trade-weighted measure of foreign inflation (Laséen and Strid, 2013; Di Casola and Iversen, 2019; Hopkins et al., 2009; Villani and Warne, 2003) Foreign policy rate: Foreign short￾term interest rate (Laséen and Strid, 2013; Di Casola and Iversen, 2019; Hopkins et al., 2009; Villani and Warne, 2003) United Kingdom CPI (Cesa-Bianchi et al., 2020; Ellis et al., 2014; Brignone and Piffer, 2025) RPIX12m inflation (Cloyne and Hurtgen, 2016) Bank Rate (Cloyne and Hurt￾gen, 2016; Elis et al., 2014; Brignone and Piffer, 2025) 1-year nominal gilt yield (Cesa-Bianchi et al., 2020) Foreign GDP: Trade-weighted World GDP (Brignone and Piffer, 2025) Foreign inflation: Trade-weighted World CPI (Brignone and Piffer, 2025)

APPENDIX 57 Table B3. Relative historical average trade weights of the US and euro area for a selection of small open economies (per cent) Country United States Euro area Canada 82 18 Czech Republic 4 96 Iceland 20 80 Norway 14 86 New Zealand 57 43 Sweden 16 84 United Kingdom 27 73 Note. Trade weights refer to the average for the period 1995-2024. Data on trade weights are collected from national central banks. For some countries, trade weight data are not available for the full sample and there￾fore represent an average based on a subset of years. Time-varying trade weights were used for all countries except Poland and the Czech Republic. Sources: National central banks and the Riksbank. APPENDIX C – Reporting of impulse response functions Here we provide an exact description of the impulse response statistics reported in the paper. These statistics are computed based on the posterior distribution of model parameters, which is represented by a sample of parameter draws from the posterior distribution obtained using the Gibbs sampler. In Figure 1, we report the posterior median impulse responses across the 10 coun￾tries. For each country, we first normalise each IRF posterior draw such that the im￾pact response of the policy rate increases by 1 percentage point on impact. Hence￾forth, it is implicitly understood that all IRFs are normalised. We then take the median across these normalised IRF posterior draws to obtain the median IRFs for each varia￾ble and country (i.e. the median IRF based on the posterior draws). Finally, we com￾pute the median response for each variable across countries (i.e. median of country median IRFs). In Figure 2, our focus is on the median peak effects for a set of variables in each coun￾try. For a given variable and country, we compute the peak value for each normalised posterior IRF draw. We then report the median peak effect (the median across poste￾rior draws) and finally compute the cross-country median peak effects. Comparing Figure 1 and 2, it may appear that the aggregate peak effects reported in the two figures are inconsistent. For example, the aggregate peak effect for GDP is −0.8 in Figure 2 but appears to be around −0.5 based on Figure 1. What explains this stark difference? We note that the posterior distribution of the IRFs includes uncer￾tainty about the timing of the peak effect. In Figure 2, we compute the distribution of the peak effect while disregarding its timing, which naturally yields a larger estimate of the peak effect.

APPENDIX 58 Similar discrepancies arise when comparing the peak effects of variables in Figure 2 with those in Figure D1, or in Figure D2 with Figure D3. Figure 2 and D3 present the distribution of the peak effect while disregarding its timing, whereas Figure D1 and D2 report the peak effect at each time period. APPENDIX D – Additional results Figure D1. Impulse responses to a monetary policy shock in ten SOEs identified using short-run zero and sign restrictions Note. The figure reports the impulse responses of a 1-percentage-point increase in the policy rate in ten small open economies identified using sign restrictions. The country models are estimated over the period 1995Q1-2024Q4, except for Chile, whose sample begins in 1995Q3. The policy rate, GDP, the unemployment rate and the real exchange rate are measured in levels while inflation is in an￾nual percentage changes.

APPENDIX 59 Figure D2. Impulse responses to a monetary policy shock in 10 SOEs identified using short-run zero restrictions Note. The figure reports the impulse responses of a 1-percentage-point increase in the policy rate in ten small open economies identified using short-run zero identification. The country models are esti￾mated over the period 1995Q1-2024Q4, except for Chile, whose sample begins in 1995Q3. The pol￾icy rate, GDP, the unemployment rate and the real exchange rate are measured in levels while infla￾tion is in annual percentage changes.

APPENDIX 60 Figure D3. Peak effects of monetary policy in ten SOEs using short-run zero restrictions Note. The figure shows the country peak impulse responses for four different variables identified using short-run zero restrictions. GDP, the unemployment rate and the real exchange rate are in lev￾els, whereas inflation is reported in annual percentage changes. Here, the peak responses are de￾fined as the lowest values along the impulse response paths for GDP, the real exchange rate and in￾flation, and as the highest values for the unemployment rate. The horizontal line refers to the me￾dian of the country-specific median peak impulse responses.

APPENDIX 61 Figure D4. Cumulative exchange rate pass-through in ten SOEs Note. The figure reports the cumulative exchange rate pass-through, calculated as the ratio of cu￾mulative inflation responses and the peak nominal exchange rate, over horizons of two, three and five years across ten small open economies. The results at the three-year horizon are shown in Fig￾ures 3 and 4. The horizontal line refers to the median of the country-specific median peak impulse responses. APPENDIX E – Regression estimates of monetary policy transmission elasticities To calculate the unconditional monetary policy transmission elasticities in our sample of SOEs, we use standard model specifications that are deliberately straightforward, mirroring the simplicity of our VAR specification. The simplicity is reflected in the fact that these models are commonly used as benchmark models in the existing literature, linear and estimated using Ordinary Least Squares (OLS). To ensure comparability, we estimate the regressions over the same sample period used in our main VAR analysis. The data employed in these regressions are identical to that used in the VAR, with the addition of a new variable, the nominal exchange rate, in the estimation of exchange rate pass-through. The estimated unconditional elasticity, 𝜇, is presented in Figure 4 and can be read from the y-axis of each subpanel.

APPENDIX 62 Okun coefficient We estimate the Okun coefficient in each country using the following specification: (𝑢 − 𝑢̅)𝑡 = 𝛼 + ∑𝛽𝑘 (𝑦 − 𝑦̅)𝑡−𝑘 2 𝑘=0

• 𝛾(𝑢 − 𝑢̅)𝑡−1 + 𝜀𝑡 (1) where (𝑢 − 𝑢̅)𝑡 and (𝑦 − 𝑦̅)𝑡 are the quarterly unemployment and GDP gaps, respec￾tively, both measured using the Hodrick-Prescott filter. The Okun coefficient is given by 𝜌 = ∑ 𝛽𝑘 2 𝑘=0 , while the long-run coefficient is calculated as 𝜇 = 𝜌 1−𝛾 . Phillips slope Our estimation of the Phillips slope for each country is guided by the model formu￾lated by Ball and Mazumder (2020), which we implement empirically as: Δ𝑝𝑡 = 𝛼 + 𝛽Δ𝑝𝑡−1 + 𝛾(𝑦 − 𝑦̅)𝑡 + 𝜀𝑡 (2) where 𝑝𝑡 is domestic consumer prices, (𝑦 − 𝑦̅)𝑡 is the quarterly GDP gap measured by the Hodrick-Prescott Filter and Δ is the difference operator. The Phillips slope coeffi￾cient is given by 𝛾 and the long-run coefficient is calculated as 𝜇 = 𝛾 1−𝛽 . Changes in domestic prices are expressed as annual quarterly percentage changes. Exchange rate pass-through For exchange rate pass-through (ERPT) in each country, we estimate the model speci￾fied by Anderl and Caporale (2023) which takes the following form: Δ𝑝𝑡 = 𝛼 + 𝛽Δ𝑝𝑡−1 + 𝛾Δ𝑁𝐸𝑅𝑡 + 𝛿Δ𝑅𝐸𝑅𝑡 + 𝜃Δ𝑝𝑡 ∗ + 𝜆(𝑦 − 𝑦̅)𝑡 + 𝜀𝑡 (3) where 𝑝𝑡 is domestic consumer prices, 𝑁𝐸𝑅𝑡 is the nominal effective exchange rate, 𝑅𝐸𝑅𝑡 is the real effective exchange rate, 𝑝𝑡 ∗ is the trade-weighted foreign consumer prices, (𝑦 − 𝑦̅)𝑡 is the GDP gap measured by the Hodrick-Prescott Filter and Δ is the difference operator. The ERPT coefficient is given by 𝛾 and the long-run coefficient is calculated as μ = γ/(1 − β). The nominal exchange rate is sourced from Bank of International Settlements (BIS) and represents the broad index, similar to the real exchange rate used in our main analysis. For all countries except Poland, we use the trade-weighted consumer prices as the measure of foreign prices. 57 Changes in prices and exchange rates are ex￾pressed as annual percentage changes. 57 See Table B3 in Appendix B for more details.

APPENDIX 63 Table E1. Comparisons of conditional monetary policy transmission elasticities calculated using different methods Okun coefficient Phillips slope Sacrifice ratio Exchange rate pass-through Peak 0.6 0.7 - - Cumulative, horizon 8 0.4 0.6 1.4 0.1 12 0.5 0.6 1.6 0.1 20 0.5 0.5 2.1 0.2 Note: Cross-country medians of cumulative monetary policy–conditional elasticities computed at 8, 12, and 20 quarters, corresponding to two-, three-, and five-year horizons.

SVERIGES RIKSBANK Tel +46 8 - 787 00 00 registratorn@riksbank.se www.riksbank.se PRODUCTION SVERIGES RIKSBANK