## UNPACKING THE EFFECTS OF TRADE POLICY UNCERTAINTY ON ASEAN ECONOMIES

## Source details

**Canonical URL:** [UNPACKING THE EFFECTS OF TRADE POLICY UNCERTAINTY ON ASEAN ECONOMIES](https://www.imf.org/-/media/files/publications/cr/2026/english/1mysea2026002-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2026/english/1mysea2026002-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2026/english/1mysea2026002-source-pdf.pdf.json)

---

### Key findings and conclusions
- Third-party protectionist TPU shocks are the most detrimental to output per capita, followed by third-party agreement TPU shocks and own protectionist TPU shocks.
- For third-party TPU shocks, declines in investment and private consumption disproportionately offset stronger trade balances that likely stem from reallocated trade flows to ASEAN.
- Own protectionist TPU shocks weigh directly on economies’ trade balances while putting upward pressure on inflation over the medium term.
- Own agreement TPU shocks are associated with a small boost to output per capita; this may reflect anticipatory investment effects for potential expanded economic activity once the tension is resolved, but the initial boost to investment and its effect on output per capita wane over time.
- TPU has reached record highs in recent years, with major protectionist peaks including: the threat of 1995 US-China trade tensions; 2018-19 US-China trade tensions; European multifront trade tensions during 2021-22; and the spike in protectionism in 2025.
- Major trade agreement TPU shocks include: Japanese auto trade talks with the US and EU in 1995; uncertainties related to China’s accession to the WTO in the early 2000s; the renegotiation of NAFTA and the formation of CP-TPP in 2018; 2019 EU-MERCOSUR negotiations; 2018-19 US-China Phase 1 trade deal; 2020 Brexit trade deal talks; and recent uncertainties around the 2025 trade negotiations between the US and its trading partners.

### Data, measurement, and methodological innovations
- Granular TPU measures constructed monthly from January 1995 to August 2025 for ASEAN6 and for 19 trading partners to build third-party measures.
- Focus economies (ASEAN6): Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam.
- Full sample of 25 economies with own TPU measures: Australia, Brazil, Canada, the Chinese mainland, Germany, Spain, France, United Kingdom, Hong Kong SAR, Indonesia, India, Italy, Japan, Korea, Mexico, Malaysia, New Zealand, Philippines, Russia, Singapore, Thailand, Taiwan Province of China, United States, Vietnam, and South Africa.
- For each of the 25 economies, three Factiva search algorithms run: country-specific protectionist TPU; country-specific agreement TPU; country-specific normalization to control for source coverage and media attention differences.
- TPU search criteria require:
  - (1) mention of a trade policy type within eight words of uncertainty and its synonyms;
  - (2) that phrase within eight words of a country reference;
  - (3) exclusion of noisy terms/phrases; and
  - (4) article length over 99 words.
- Protectionist term examples: tariffs, trade restrictions, trade war, protectionism, export licenses, import licenses, local content requirements.
- Agreement term examples: World Trade Organization, WTO, FTA, trade deal, trade agreement, trade negotiations.
- Raw TPU measures = counts of matching articles divided by a 13-month centered moving average of a normalization count; values bounded between 0 and 1 and interpretable as share of country media coverage related to the TPU type.
- The 50 raw measures (25 protectionist TPU and 25 agreement TPU) were scaled to have a mean of 100 over 1995 to 2015 across both TPU types; e.g., a value of 150 implies 50 percent higher than the 1995-2015 historical average.
- Third-party TPU indices for each ASEAN6 country are weighted averages of trading partners’ own TPU indices using each ASEAN6 country i’s export shares to destination j; annual export shares through 2024 used to smooth monthly volatility.
- On average, the share of total average exports covered by the 24 j economies used in the third-party index is 85 percent.
- Export coverage by ASEAN6 (average over sample): Indonesia (86.2 percent); Malaysia (86.6 percent); Philippines (89.5 percent); Singapore (84.6 percent); Thailand (79.0 percent); Vietnam (83.5 percent).

### Stylized facts and sample statistics
- ASEAN6’s own protectionist and own agreement average TPU is low and less volatile compared to other regions; other Asia (largely driven by China and Japan) has the highest averages and standard deviations across both TPU types.
- For almost all ASEAN6 economies, both own protectionist and agreement TPU spiked in 2025 following higher US tariffs and associated uncertainty with trade deals.
- Malaysia-specific TPU drivers: protectionist TPU rose following a 2012 US anti-dumping threat on magnesium exports, uncertainties related to Malaysia’s response to the 2018 US-China tensions, and a 2022 food export ban; agreement TPU peaked following 2007 Malaysia-US FTA talks and a threatened 2019 WTO palm oil dispute with the EU.

### Empirical methodology (specification and identification)
- Sample and frequency: Annual data from 1995 to 2024; horizons h = 0 through 5 (up to five years ahead).
- Estimation approach: Jorda local projections tracing cumulative effects of TPU shocks; four baseline regressions for Own Protectionist TPU, Own Agreement TPU, Third-Party Protectionist TPU, Third-Party Agreement TPU.
- TPU shocks defined as dummies that take value 1 if TPU is at least one standard deviation above the ASEAN6 mean; continuous standardized shocks used as robustness.
- Controls and selection:
  - y_{i,t-1} included to account for persistence.
  - X_{i,t} vector of controls selected using post-double-selection LASSO on over 150 macro-financial controls.
  - FPEP (revision control) = control for revisions to IMF growth forecasts between Spring and Fall WEOs in year t-1.
  - θ_{i,h} are country fixed effects; time fixed effects not included for third-party shocks.
- Inference: wild-cluster bootstrap procedure used to address small-sample bias; 90% wild-bootstrap CIs reported in robustness figures.
- Caveats: TPU measures non-normal; impulse responses often not statistically significant at most horizons; TPU indices may partially capture media attention shifts; regressions abstract from general equilibrium feedbacks.

### Empirical results — headline impacts on real GDP per capita
- Own protectionist TPU shocks:
  - Statistically significant cumulative decline in real GDP per capita of 3.7 percent in period t+1 relative to pre-shock and counterfactual.
  - In dual-shock specification (own and third-party together), own protectionist TPU effect falls to a cumulative decline of 3.0 percent in period t+1.
- Own agreement TPU shocks:
  - Positive significant average effect on real GDP per capita of 1.3 percent in the initial period.
- Third-party TPU shocks (medium term):
  - Third-party protectionist TPU shocks associated with a 7.1 percent cumulative decline in real GDP per capita by period t+2.
  - Third-party agreement TPU shocks associated with a cumulative 5.7 percent decline in real GDP per capita by period t+2.
- Context for magnitudes:
  - Over the sample, average growth in real GDP per capita across the ASEAN6 is 3.1 percent per annum and median growth is 3.8 percent.
  - Continuous-shock robustness: an increase in third-party protectionist TPU of magnitude comparable to the 2018 China-U.S. trade tensions is associated with a cumulative decline in GDP per capita of 8.7 percent at t+2.

### Trade openness interaction and heterogeneity
- Interaction result: coefficient on interaction between third-party TPU shocks and average trade openness is positive and significant in period t+2.
- Percentile effects:
  - At the 10th percentile of trade openness, third-party TPU shocks lead to declines in output per capita of around 9 percent by t+2.
  - At the 90th percentile of openness (e.g., Singapore), the effect of third-party TPU shocks becomes insignificant.
- Interpretation: higher trade openness helps offset negative effects of third-party TPU shocks by enabling rapid reallocation across sources of demand and supply.

### Channels of impact (selected outcomes and magnitudes)
- Own protectionist TPU shocks:
  - Exports to GDP fall significantly by 1.9 percentage points in the initial year of the shock.
  - Investment to GDP and net FDI inflows to GDP cumulatively increase by 3.6 percentage points by year t+2.
  - Inflation cumulatively increases by 3 percentage points by year t+3.
  - Private consumption: positive impact reported (possible confidence effect from anticipated investment).
  - Current account balance: negative over time (not precisely estimated).
  - Equity markets: initially fall then recover beyond t+2.
- Third-party protectionist TPU shocks:
  - Exports to GDP drift upwards over the horizon.
  - Current account balance to GDP increases by 2.5 percentage points by t+2.
  - Investment and private consumption decline, offsetting stronger trade balances.
  - Equity markets: temporarily respond positively on average.

### Malaysia: protectionist and agreement TPU shocks (country-specific outcomes)
- Own protectionist TPU shocks:
  - Negatively affect growth via worsened trade balances and push up inflation over the medium term.
- Third-party protectionist TPU shocks:
  - More detrimental to output than own protectionist shocks; driven by investment and consumption declines that outweigh trade balance improvements.
- Own agreement TPU shocks:
  - Initially support output through higher investment in anticipation of stronger activity; the boost wanes over time.
  - Do not lead to sustained improvements in trade or consumption; current account balances fall an average of 2.0 percentage points in the initial year as imports rise more than exports.
- Quantitative impacts (Malaysia-specific reported figures):
  - Protectionist TPU shocks: investment and private consumption ratios cumulatively fall by 4.2 and 1.3 percentage points by t+2 and t+1, respectively.
  - Third-party agreement TPU shocks: investment to GDP ratio falls by 2.4 percentage points by t+2; current account balance improves by 1.4 percentage points in t+1.
  - Own agreement TPU shocks: investment and FDI to GDP ratios rise early by around 2 percentage points of GDP; current account balances fall an average of 2.0 percentage points in the initial year.

### Policy implications and recommendations
- Maintain predictable trade policy frameworks to sustain investment, stability, and long-term growth amid global trade tensions.
- Be cautious in interpreting short-term trade gains as net benefits—evaluate broader macroeconomic channels (investment, consumption, inflation).
- Consider trade openness and regional integration/diversification as policy levers to mitigate TPU shocks.
- For Malaysia and similar economies: avoid over-reliance on temporary diversion gains; prioritize policies that protect investment and consumption during global TPU episodes.

### Robustness checks, extensions, and suggested next steps
- Robustness checks performed:
  - Adding controls for historical crises (Asian Financial Crisis, Global Financial Crisis, Covid-19 pandemic) — results qualitatively similar.
  - Including all four TPU shocks in a single regression — qualitatively similar though less precise.
  - Using standardized continuous versions of TPU measures (one standard deviation increases) — qualitatively similar.
  - Inclusion of time fixed effects for own-country TPU shocks and allowing lags of shock variables — qualitatively robust.
- Precision caveats: impulse responses often not statistically significant at most horizons; smaller panel and potential omitted global factors limit causal interpretation.
- Suggested extensions:
  - Further refine TPU algorithms and try alternative shock magnitudes and placebo tests.
  - Use financial openness and regional measures of trade integration or diversification as alternative interaction terms.
  - Explore higher-frequency specifications (quarterly and monthly) for inflation and financial market outcomes.
  - Adopt a firm-level framework to uncover microeconomic responses, increase statistical power, and identify mechanisms driving aggregate responses.

### Data sources and variable coverage (selected list)
- Core macro series: WEO Database (real growth, inflation, exports, imports, current account balance, investment, private consumption, public consumption, GDP per capita, population, historical WEO vintages).
- FDI: World Bank.
- Stock market indices, short-term rates, real effective exchange rate, and money supply: CEIC; Haver; Bloomberg.
- Natural resource rents, institutional quality, trade openness, effective tariff rate: World Bank; WTO.
- Financial development index: IMF.
- VIX and US stock market index: Bloomberg.

*Source: IMF staff calculations and analysis in the provided chapter content.*

### References ____________________________________________________________________________ 20

### UNPACKING THE EFFECTS OF TRADE POLICY UNCERTAINTY ON ASEAN ECONOMIES

### Key findings and conclusions
- Third-party protectionist TPU shocks are the most detrimental to output per capita, followed by third-party agreement TPU shocks and own protectionist TPU shocks.
- For third-party TPU shocks, declines in investment and private consumption disproportionately offset stronger trade balances that likely stem from reallocated trade flows to ASEAN.
- Own protectionist TPU shocks weigh directly on economies’ trade balances while putting upward pressure on inflation over the medium term.
- Own agreement TPU shocks are associated with a small boost to output per capita; this may reflect anticipatory investment effects for potential expanded economic activity once the tension is resolved, but the initial boost to investment and its effect on output per capita wane over time.
- TPU has reached record highs in recent years, with major protectionist peaks including: the threat of 1995 US-China trade tensions; 2018-19 US-China trade tensions; European multifront trade tensions during 2021-22 (including post-Brexit UK disputes and steel trade tensions); and the spike in protectionism in 2025.
- Major trade agreement TPU shocks include: Japanese auto trade talks with the US and EU in 1995; uncertainties related to China’s accession to the WTO in the early 2000s; the renegotiation of NAFTA and the formation of CP-TPP in 2018; 2019 EU-MERCOSUR negotiations; 2018-19 US-China Phase 1 trade deal; 2020 Brexit trade deal talks; and recent uncertainties around the 2025 trade negotiations between the US and its trading partners.

### Data, measurement, and methodological innovations
- Granular TPU measures were constructed for the ASEAN6 at a monthly frequency from January 1995 to August 2025, and for 19 trading partners to construct third-party measures.
- The analysis focuses on six ASEAN economies (ASEAN6): Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam.
- The full sample of 25 economies for which own TPU measures were produced includes: Australia, Brazil, Canada, the Chinese mainland, Germany, Spain, France, United Kingdom, Hong Kong SAR, Indonesia, India, Italy, Japan, Korea, Mexico, Malaysia, New Zealand, Philippines, Russia, Singapore, Thailand, Taiwan Province of China, United States, Vietnam, and South Africa.
- For each of the 25 economies, three search algorithms were run on Dow Jones’ Factiva: a country-specific protectionist TPU algorithm, a country-specific agreement TPU algorithm, and a country-specific normalization algorithm to control for differences in Factiva’s source coverage over time and broader disparities in media attention.
- The TPU search criteria require: (1) mention of a trade policy type within eight words of uncertainty and its synonyms; (2) that phrase within eight words of a country reference; (3) exclusion of noisy terms/phrases (e.g., stock market trading); and (4) article length over 99 words.
- Protectionist term examples: tariffs, trade restrictions, trade war, protectionism, export licenses, import licenses, local content requirements.
- Agreement term examples: World Trade Organization, WTO, FTA, trade deal, trade agreement, trade negotiations.
- Raw TPU measures are counts of matching articles divided by a 13-month centered moving average of a normalization count, producing values bounded between 0 and 1 interpretable as the share of country media coverage related to the TPU type.
- The 50 raw measures (25 protectionist TPU and 25 agreement TPU) were scaled to have a mean of 100 over the period 1995 to 2015 across both TPU types; e.g., a value of 150 implies 50 percent higher than the 1995-2015 historical average.
- Third-party TPU indices for each ASEAN6 country are weighted averages of trading partners’ own TPU indices using each ASEAN6 country i’s export shares to destination j as weights. Annual export shares through 2024 are used to smooth monthly volatility.
- On average, the share of total average exports covered by the 24 j economies used in the third-party index is 85 percent.
- Export coverage by ASEAN6 country (average over the sample period) is: Indonesia (86.2 percent); Malaysia (86.6 percent); Philippines (89.5 percent); Singapore (84.6 percent); Thailand (79.0 percent); Vietnam (83.5 percent).

### Stylized facts and sample statistics
- ASEAN6’s own protectionist and own agreement average TPU is low and less volatile compared to other regions; other Asia (largely driven by China and Japan) has the highest averages and standard deviations across both TPU types.
- For almost all ASEAN6 economies, both own protectionist and agreement TPU spiked in 2025 following higher US tariffs and associated uncertainty with trade deals.
- Malaysia-specific TPU drivers: protectionist TPU rose following a 2012 US anti-dumping threat on magnesium exports, uncertainties related to Malaysia’s response to the 2018 US-China tensions, and a 2022 food export ban; agreement TPU peaked following 2007 Malaysia-US FTA talks and a threatened 2019 WTO palm oil dispute with the EU.

### Empirical implications and channels
- Third-party protectionist shocks reduce real GDP per capita most strongly via declines in investment and private consumption, despite improved trade balances that likely reflect reallocated trade flows to ASEAN.
- Own protectionist shocks directly harm trade balances and put upward pressure on inflation over the medium term.
- Own agreement shocks can temporarily boost investment and output per capita (anticipatory effects), but these effects dissipate and do not translate into stronger trade balances.

_References and figures/tables cited are from the IMF staff analysis contained in the source document._

### 9.      The third-party protectionist and agreement TPU measures follow broadly the same

### 9. The third-party protectionist and agreement TPU measures follow broadly the same pattern across ASEAN6 countries

### Key historical TPU spikes and patterns
- Protectionist third-party TPU spiked following:
  - 1995 US-Japan auto trade negotiations / US-China trade tensions in 1995
  - 2018-19 US-China trade tensions
  - 2024 trade disputes between China and some of its trading partners (US, Canada, and the EU)
- Third-party agreement TPU spiked following:
  - US-Japan auto trade negotiations in 1995
  - China’s WTO accession in the early 2000s
  - 2017-18 NAFTA renegotiation and the formation of CP-TPP
  - 2019 US-China Phase 1 deal
- Figure coverage: 1995 to 2024 (Figure 3 does not include 2025, as full-year trade data were not yet available).

### Correlation of granular TPU measures with existing uncertainty measures
- ASEAN6 protectionist TPU correlation coefficients with alternative measures:
  - Global EPU (Baker, Bloom, and Davis, 2016): 0.24
  - US TPU (Baker, Bloom, and Davis, 2016): 0.37
  - Caldara et al. (2019) TPU: 0.37
  - World Trade Uncertainty Index (WTUI) (Ahir et al., 2022): 0.49
  - China TPU: 0.30
  - Equity Market Volatility TPU: 0.33
- ASEAN6 agreement TPU correlations (not highly correlated except with WTUI):
  - Trade Agreement correlation shown as 0.191 in table excerpt
  - Global EPU: -0.04
  - US TPU: 0.10
  - WTUI: 0.31
  - Caldara TPU: 0.09
  - China TPU: 0.03
  - Equity Market Volatility TPU: 0.09
- Interpretation: Agreement TPU measures provide a large degree of new information not captured by alternative uncertainty measures.

### Empirical methodology
- Sample and frequency:
  - Annual data from 1995 to 2024.
  - Horizons h = 0 through 5 (up to five years ahead).
- Estimation approach:
  - Local projection specifications tracing cumulative effects of TPU shocks.
  - Four baseline regressions for Own Protectionist TPU, Own Agreement TPU, Third-Party Protectionist TPU, Third-Party Agreement TPU.
- Key variable definitions and features:
  - i = country (each of the 6 ASEAN6 countries).
  - t = year (1995 through 2024).
  - y = macroeconomic outcome of interest: log real GDP per capita; inflation; external balance ratios; investment ratios; consumption ratios; and log equity market indices.
  - TPU shocks are dummies that take a value of 1 if the associated TPU is at least one standard deviation above the ASEAN6 mean.
  - OEAO (trade openness) = sample average (1995 to 2024) of trade openness measured as sum of exports and imports as a share of GDP.
  - y_{i,t-1} included to account for persistence.
  - X_{i,t} is a vector of controls selected using the post-double-selection, machine-learning based LASSO approach on a comprehensive set of over 150 macro-financial controls.
  - FPEP (revision control) = control for revisions to IMF growth forecasts between Spring and Fall World Economic Outlooks in year t-1.
  - θ_{i,h} are country fixed effects; time fixed effects are not included (third-party shocks highly correlated across countries).
  - Standard errors computed using a wild-cluster bootstrap procedure to address small sample bias concerns.
- Caveats on TPU shocks:
  - Measures do not follow a normal distribution; dummy approach identifies exogenous and sizeable TPU spikes similar to literature (e.g., Bloom, 2009). Continuous standardized shocks used as robustness check.

### Empirical results — headline impacts on real GDP per capita
- Own protectionist TPU shocks:
  - Statistically significant cumulative decline in real GDP per capita of 3.7 percent in period t+1 relative to pre-shock and counterfactual (Figure 4, upper left).
  - In dual-shock specification (own and third-party shocks together), own protectionist TPU effect falls to a cumulative decline of 3.0 percent in period t+1.
- Own agreement TPU shocks:
  - Positive significant average effect on real GDP per capita of 1.3 percent in the initial period (Figure 4, upper right).
- Third-party TPU shocks (medium term):
  - Third-party protectionist TPU shocks associated with a 7.1 percent cumulative decline in real GDP per capita by period t+2 (Figure 4, bottom left).
  - Third-party agreement TPU shocks associated with a cumulative 5.7 percent decline in real GDP per capita by period t+2 (Figure 4, bottom right).
- Context for magnitudes:
  - Over the sample, average growth in real GDP per capita across the ASEAN6 is 3.1 percent per annum and median growth is 3.8 percent.
  - Continuous-shock robustness: an increase in third-party protectionist TPU of magnitude comparable to the 2018 China-U.S. trade tensions is associated with a cumulative decline in GDP per capita of 8.7 percent at t+2.

### Trade openness interaction and heterogeneity
- Interaction results:
  - Coefficient on interaction between third-party TPU shocks and average trade openness is positive and significant in period t+2.
  - Higher trade openness helps offset negative effects of third-party TPU shocks.
- Percentile effects (Figure 5):
  - At the 10th percentile of trade openness, third-party TPU shocks lead to declines in output per capita of around 9 percent by t+2.
  - At the 90th percentile of openness (e.g., Singapore), the effect of third-party TPU shocks becomes insignificant.
- Interpretation:
  - Trade openness can buffer economies by allowing rapid reallocation across sources of demand and supply, especially for diversified economies.

### Robustness checks and alternative specifications
- Robustness exercises reported:
  - Adding controls for historical crises (Asian Financial Crisis, Global Financial Crisis, Covid-19 pandemic) — results qualitatively similar (Annex V).
  - Including all four TPU shocks in a single regression — qualitatively similar though less precise.
  - Using standardized continuous versions of TPU measures (one standard deviation increases) — qualitatively similar.
  - Inclusion of time fixed effects for own-country TPU shocks and allowing for potential inclusion of lags of shock variables — qualitatively robust.
- Precision and interpretation caveats:
  - Impulse responses are not statistically significant at most horizons; smaller panel and potential omitted global factors limit causal interpretation.
  - TPU indices may partially capture media attention shifts rather than underlying policy uncertainty.
  - Regressions abstract from general equilibrium effects and feedback mechanisms.

### Channels of impact (selected outcomes and magnitudes)
- Own protectionist TPU shocks (Figure 7, red lines/areas):
  - Exports to GDP fall significantly by 1.9 percentage points in the initial year of the shock; negative effect builds but loses significance over time.
  - Current account balance impact: negative over time (not precisely estimated).
  - Investment to GDP and net FDI inflows to GDP cumulatively increase by 3.6 percentage points by year t+2.
  - Inflation cumulatively increases by 3 percentage points by year t+3.
  - Equity markets: initially fall then recover beyond t+2.
  - Private consumption: positive impact reported (interpretation: possibly due to signaling of forthcoming large investments buoying confidence).
- Third-party protectionist TPU shocks (Figure 7, blue lines/areas):
  - Exports to GDP drift upwards over the horizon.
  - Current account balance to GDP increases by 2.5 percentage points by t+2.
  - Negative effects on investment and private consumption offset stronger trade balances.
  - Equity markets: temporarily respond positively (average rise reported in text fragment).

*Source: IMF staff calculations and analysis in the provided chapter content.*

### 15.6 percent in the initial year. However, over time, lower investment and private consumption ratios

### Malaysia: Protectionist and Agreement TPU Shocks

### Effects on Growth and Key Channels
- Own protectionist TPU shocks:
  - Negatively affect growth via worsened trade balances.
  - Push up inflation over the medium term.
  - Aggregate negative effect on real output per capita; caution against assuming “winners” from trade tensions despite stronger trade outcomes.
- Third-party protectionist TPU shocks:
  - More detrimental to output than own protectionist shocks.
  - Driven by investment and consumption declines that outweigh improvements in trade balances.
  - May prompt precautionary savings behavior (“waiting and seeing”) despite positive effects on sectors benefitting from diversion.
- Own agreement TPU shocks:
  - Initially support output through higher investment in anticipation of stronger economic activity.
  - Temporary: the boost to investment in early years wanes over the horizon.
  - Do not lead to sustained improvements in trade or consumption; current account balances fall an average of 2.0 percentage points in the initial year as imports rise more than exports.

### Quantitative Impacts Reported
- Protectionist TPU shocks:
  - Investment and private consumption ratios cumulatively fall by 4.2 and 1.3 percentage points by t+2 and t+1, respectively.
- Third-party agreement TPU shocks:
  - Investment to GDP ratio falls by 2.4 percentage points by t+2.
  - Current account balance improves by 1.4 percentage points in t+1.
- Own agreement TPU shocks:
  - Investment and FDI to GDP ratios rise early in the horizon by around 2 percentage points of GDP.
  - Current account balances fall an average of 2.0 percentage points in the initial year.

### Trade and Openness Interaction
- Trade openness mitigates the adverse effects of TPU shocks.
- Third-party agreement TPU shocks may prompt third-party suppliers to seek alternative sources amongst ASEAN economies, contributing to trade balance improvements even as investment falls.
- Despite temporary improvements in trade outcomes for some shocks, negative effects on investment and consumption can dominate, reducing real output per capita.

### Policy Implications and Recommendations
- Maintain predictable trade policy frameworks to sustain investment, stability, and long-term growth amid global trade tensions.
- Be cautious in interpreting short-term trade gains as net benefits—evaluate broader macroeconomic channels (investment, consumption, inflation).
- Consider trade openness and regional integration/diversification as policy levers to mitigate TPU shocks.

### Methodology Highlights
- TPU algorithms constructed monthly for 25 economies from January 1995 to August 2025 using word banks based on Baker et al. (2016), journalist consultations, and country-specific Factiva filters.
- Algorithms:
  - Allow term variation (e.g., “tariffs on imports”) but exclude articles mentioning other policy uncertainty types to avoid cross-contamination.
  - Exclude references to absence of uncertainty (e.g., “without doubt”, “no uncertainty”) and equity market volatility when unrelated to policy uncertainty.
- Example search normalizer and constructs for Malaysia include:
  - Normalizer: re=MALAY and (Malaysia*) and wc>99
  - Trade II (Protectionism Uncertainty): re=MALAY and (Malaysia*) near8 (((trade) near2 (restriction or quota or tariff* or sanction or embargo or subsid* or war or barrier or red tape)) or protectionism or dumping ... ) near8 (uncert* or ambiguous or dubious or precarious or unpredictable or undecided or undetermined or unresolved or unsettled or fear or risk or threat* or doubt or worry) not (without doubt or no uncertainty ... ) and wc>99
  - Trade III (Agreement Uncertainty): re=MALAY and (Malaysia*) near8 (((trade) near2 (commission or agreement or bloc* or multilateral or negotiation* or deal or talks)) or WTO or World Trade Organization or FTA) near8 (uncert* or ambiguous ... ) not (without doubt or no uncertainty ... ) and wc>99

### Suggested Extensions and Robustness Checks
- Further refine TPU algorithms and try other alternative shock magnitudes and a placebo test.
- Use financial openness and regional measures of trade integration or diversification as alternative interaction terms.
- Explore higher-frequency specifications (quarterly and monthly) for inflation and financial market outcomes to better capture short-term effects and financial transmission.
- Adopt a firm-level framework to uncover microeconomic responses, increase statistical power, and identify mechanisms driving aggregate responses.

*Sources: IMF staff calculations; excerpt from IMF chapter on Malaysia TPU analysis.*

### Annex IV. Empirical Methodology and Data Annex

### Annex IV. Empirical Methodology and Data Annex

### Methodology: Local Projections and Identification
- Jorda local projections are used to estimate horizon-specific responses without imposing strong structural assumptions (e.g., relative to a vector autoregressive approach).
- In addition to baseline controls, a post-double selection LASSO machine-learning approach (Belloni et al., 2014) is employed to select additional controls given model uncertainty and the small size of the panel.
- Two LASSO regressions are estimated for variable selection: one of the dependent variable on all candidate controls, and another of the shock variable on the same set. The union of selected variables forms the final control set.
- To avoid mechanical correlations across related aggregates, the LASSO estimation excludes variables that are direct components or close accounting counterparts of the dependent variable (examples provided in the source text).

### Variable Selection and Functional Forms
- The candidate control vector, X_it, includes a broad range of macroeconomic, financial, and structural factors:
  - real growth, inflation, trading partner real growth, exports, imports, financial development, fiscal balances, investment, population, private and public consumption, unemployment, broad money, natural resource rents, exchange rate volatility, stock market indices, the real effective exchange rate, tariff rates, government revenue and spending ratios, national saving, debt ratios, investment and FDI ratios, current account balances, oil prices, the VIX, the S&P US equity market index, and a Covid-period dummy.
- Transformations of each control variable are also included:
  - first differences, lags, lagged differences, and squared terms to capture dynamic and nonlinear relationships.
- The list of variables selected by the procedure is broadly consistent with economic priors. Example (for real GDP per capita):
  - financial development (lagged, and squared terms),
  - foreign direct investment (current and lagged),
  - the effective tariff rate,
  - imports,
  - lagged private consumption,
  - lagged national saving,
  - fiscal indicators such as the public debt ratio squared and lagged government spending.

### Inference and Standard Errors
- Standard errors are based on a wild-cluster bootstrap-t procedure that produces robust inference even with few clusters (Cameron, Gelbach, and Miller, 2008).
- Bootstrap confidence intervals referenced in robustness figures are 90% wild-bootstrap CIs; clustered by id (Webb weights) or clustered by id & year depending on the figure.

### Data Sources (variables, frequencies, and source institutions)
- Real Growth (y-o-y percent change): Annual, Quarterly — WEO Database
- Inflation (y-o-y percent change): Average, Quarterly — WEO Database
- Exports: Annual — WEO Database
- Imports: Annual — WEO Database
- Current Account Balance (percent of GDP): Annual — WEO Database
- Investment (percent of GDP): Annual — WEO Database
- Foreign direct investment (percent of GDP): Annual — World Bank
- Private Consumption (percent of GDP): Annual — WEO Database
- Public Consumption (percent of GDP): Annual — WEO Database
- Unemployment: Annual — WEO Database
- Stock Market Index: Annual, Monthly — CEIC
- 10Y Sovereign Bond Spreads: Annual, Quarterly, Monthly — Bloomberg
- Corporate Bond Spreads: Annual, Quarterly, Monthly — Bloomberg
- GDP per capita: Annual — WEO Database
- Education: Annual — World Bank
- Population level: Annual — WEO Database
- Trade Openness: Annual — World Bank
- Institutional Quality: Annual — World Bank
- Private Credit to GDP: Annual, Quarterly — BIS
- Financial Development Index: Annual — IMF
- Natural Resource Rents (percent of GDP): Annual — World Bank
- Money Supply Growth (broad money): Annual, Monthly — CEIC
- Exchange Rate (to USD): Annual, Quarterly, Monthly — CEIC
- Exchange Rate Volatility: Annual, Quarterly, Monthly — Haver
- Real Effective Exchange Rate: Annual, Monthly — CEIC
- Fiscal Overall Balance: Annual — WEO Database
- Inflation Expectations: Annual, Monthly — Ecdatabase
- Effective Tariff Rate: Annual — World Bank, World Trade Organization
- Oil Prices: Annual, Quarterly, Monthly — Haver
- Port Quality: Annual — CEIC
- Historical WEO Vintages: Annual — IMF
- Short-Term Policy Rates: Annual, Quarterly, Monthly — Haver
- Short-Term Interest Rates: Annual — Haver
- VIX: Annual, Quarterly, Monthly — Bloomberg
- US Stock Market Index: Annual, Quarterly, Monthly — Bloomberg

### Robustness Checks (Annex V excerpts)
- Robustness specifications include dummy controls for global shocks: Covid-19 Pandemic (2020-22), Global Financial Crisis (2008-09), Asian Financial Crisis (1997-98), and combinations (e.g., All Three Controls).
- Alternative specifications present:
  - inclusion of all four shocks in the same regression,
  - continuous-shock specifications using standardized z-scores of TPU measures,
  - 1-s.d. shock responses for Own Protectionism TPU, 3rd Party Protectionism TPU (Avg), Own Agreement TPU, and 3rd Party Agreement TPU (Avg).
- Figures report impulse responses of Real GDP Per Capita (log) to TPU shocks with y-axis labeled in Response (percent) and x-axis in Years after shock up to 5 years; shaded areas denote 90% wild-bootstrap CIs.
- One multi-shock figure plots percent responses with series: Own Protectionism TPU, Own Agreement TPU, 3rd Party Protectionism TPU, 3rd Party Agreement TPU; CIs clustered by id & year.

*Source: Annex IV. Empirical Methodology and Data Annex (as supplied).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2026/english/1mysea2026002-source-pdf.pdf_