## 1. Aggregate Demand ## Source details **Canonical URL:** [1. Aggregate Demand](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1630.pdf) ## Other formats - [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1630.pdf.md) - [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1630.pdf.json) --- ### I. Research question and empirical setting - Research objective: empirically investigate whether the global financial crisis induced heterogeneous effects on traded goods differentiated by quality (a “flight from quality”). - Contextual facts: - Great Trade Collapse: world trade fell by 30 percent in nominal terms, and 18 percent in real terms between 2008Q3 and 2009Q2. - Crisis period for Argentine wine defined as 2008Q4 to 2009Q3. - Data: - Customs dataset (2002–2009) with firm name, destination country, date of shipment, FOB value (US$), and volume (liters); wine identified by name, grape, type, and vintage year (vintage ignored for main product definition). - Quality measures: Wine Spectator and Robert Parker scores on a (50,100) scale; product quality computed as unweighted average across same name, grape, and type (weighted averages used in robustness). - Price proxy: unit values = value (US$) / volume (liters); exports are FOB. - Quarterly aggregation and filters: shipments smaller than 4.5 liters excluded; only wine producers retained; main sample span 2006Q4 to 2009Q3. - Sample composition (matched Wine Spectator / Parker where noted): - Matched Wine Spectator sample: 138 firms, 1,167 wines, 437 different names, 3 types, 22 grapes; Wine Spectator score range: 68 to 96. - Matched Parker sample: 107 firms, 595 wines, 246 different names, 3 types, 18 grapes; Parker score range: 74 to 98. - Full sample: 138 exporters, 1,167 wines, 84 destination countries, 8,051 observations. - Pre-crisis sample: 126 firms, 996 wines, 81 countries, 4,344 observations. - Crisis sample: 115 firms, 945 wines, 76 countries, 3,707 observations. ### II. Main empirical findings: flight from quality - Direction and magnitude: - Before the crisis: a one unit increase on the quality scale raised export growth by 2 percentage points on average. - During the crisis: a one unit increase on the quality scale reduced export growth by 2 percentage points on average. - Baseline regression (Panel A, Table 8): quality × D_crisis = −0.036 (0.013) for export values; quality × D_crisis = −0.031 (0.013) for export volumes; quality × D_crisis = −0.005 (0.003) for unit values (weak significance). - Aggregate crisis movements (samples including 2,247 and 1,167 wines): - Nominal exports fell by 29.48 and 36.53 percent, respectively, during the crisis. - Real exports fell by 32.37 and 38.17 percent, respectively. - Unit values rose by 4.11 and 2.57 percent, respectively. - Counterfactuals on quality composition: - If quality had increased to maximum observed (96): nominal exports would have dropped by 38.94 percent (compared to actual fall of 36.53 percent). - If quality had fallen to minimum observed (68): nominal exports would have fallen by 30.00 percent. - Difference in nominal trade performance due to quality composition during the crisis: 9 percentage points. - For real exports: difference across extreme quality scenarios ≈ 8 percentage points (40.63 percent vs 32.39 percent; actual decrease 38.17 percent). ### III. Intensive vs extensive margin decomposition - Aggregate change in nominal wine exports during crisis: −29.48 percent (from 178,071 to 125,585 thousand US dollars). - Extensive-margin changes: - Number of exporters fell by 5.33 percent (∆f = 0.9467). - Mean number of destinations fell by 6.84 percent (∆d = 0.9316). - Mean number of wines per firm-destination fell by 1.71 percent (∆p = 0.9829). - Combined extensive-margin contribution: −13.31 percent. - Intensive margin: - Mean value per firm-destination-product fell by 18.65 percent (∆x = 0.8135). - Intensive margin contribution: 63.27 percent of the collapse; extensive margin contribution: 36.73 percent. - Intensive-margin components: mean quantities fell by 21.99 percent (∆q = 0.7801); mean unit values rose by 4.28 percent (∆uv = 1.0428). - Regression focus: intensive-margin changes (difference-in-difference on firm-product-destination triplets reporting positive exports in consecutive periods) and robustness including extensive-margin specifications. ### IV. Quantity vs price channel - Collapse in nominal exports for higher-quality wines driven essentially by a drop in quantities rather than prices (unit values interactions with crisis mostly insignificant). - Intensive-margin breakdown shows quantities fell substantially while unit values increased overall, but unit-value interactions with quality were generally small. ### V. Heterogeneity by destination and market characteristics - Country-specific crisis interactions (selected estimates): - United States: quality × D_crisis × US on export values = −0.112 (0.023). - United Kingdom: quality × D_crisis × UK on export values = −0.073 (0.023). - China: interactions for nominal and real exports insignificant; some unit value effects. - Hong Kong: quality × D_crisis × Hong Kong raised nominal and real export growth by 7.0 and 7.8 percentage points respectively (liberalization measures noted). - Wine-producing vs non-producing and self-sufficiency: - Quality × crisis interaction for five major producers = −0.087 (Table 12, Panel A, column (1)). - Self-sufficiency interactions (Table 13): quality × crisis = −0.073 for self-sufficient destinations; −0.026 for non-self-sufficient producers; insignificant for non-producers in many specifications. - Interpretation: flight from quality stronger in markets where domestic wine production/self-sufficiency allows substitution toward domestic varieties. ### VI. Firm size heterogeneity - Firm-size classification: small = smallest 80 percent of exporters by volume; large = top 20 percent. - Findings (Table 15): - Exports of higher-quality wines collapsed more during the crisis for small firms (quality × D_crisis × small on export values = −0.048 (0.019)); no significant effect for large firms. - Results for unit values remain mostly insignificant. - Domestic interbank lending rate rose from an average of 8.8 to 10.9 percent between 2007Q4–2008Q3 and 2008Q4–2009Q3, consistent with tighter credit conditions affecting small firms. ### VII. Nonlinearities, extensive margin inclusion, and post-crisis recovery - Nonlinear quality effects: - Semi-elasticities rise with quality pre-crisis and become more negative at higher quality during crisis. - Examples (semi-elasticities for nominal export growth): - Pre-crisis: “Good” 0.015; “Very good” 0.030; “Outstanding” 0.045; “Great” 0.061. - Crisis: “Good” −0.014; “Very good” −0.033; “Outstanding” −0.051; “Great” −0.072. - Extensive margin methods: - Alternate dependent variables and estimators (mid-point growth, ln(1+X), Tobit, Poisson PML) broadly corroborate baseline negative quality × D_crisis effects, though magnitudes vary. - Post-crisis recovery (comparison between 2008Q4–2009Q3 and 2011Q4–2012Q3 due to missing 2010 data): - Estimated α1 (pre-crisis effect), α2 (crisis differential), α3 (post-crisis differential): α1 > 0, α2 < 0, α3 > 0 in main specifications. - Example (Table 18, Panel A, column 1): quality × D_post implies one unit increase in quality raised export growth by 6.3 percentage points net over the period (0.018 + 0.045). - Conclusion: flight from quality was temporary; higher-quality exports recovered more strongly after the crisis. ### VIII. Channels explored - Aggregate demand channel: - Country-level mean changes 2007–2009: real GDP per capita growth from 1.6 to −3.9 percent; household expenditures per capita growth from 1.6 to −2.0 percent; unemployment rate change from −0.1 to 1.9 percent. - Interaction of quality with income growth positive and significant: ∆lnAD × quality on export values = 0.411 (0.139). - Interpretation: destination aggregate-demand declines disproportionately reduced higher-quality exports. - Expenditure switching: - Stronger flight from quality when destination countries are wine producing and self-sufficient; largest producers saw steepest declines in higher quality exports (e.g., −0.087 for five largest producers). - Inventory adjustments: - Change in shipment frequency positively associated with export values; interaction with quality generally insignificant for values and volumes, suggesting inventory depletion affects trade flows but does not explain flight-from-quality heterogeneity. - Firm-size and credit constraints: - Small firms exhibited stronger flight-from-quality patterns, consistent with demand sensitivity and tighter finance during crisis. - Exchange-rate interactions: - Estimated but coefficients insignificant (limited time-series variation). ### IX. Robustness and endogeneity checks - Instrumental-variable approach: - Parker scores used as instrument for Wine Spectator ratings; Kleibergen-Paap F statistic = 437 (critical value = 16), first-stage coefficient = 0.703 (significant at 1%). - IV estimates increase magnitude for export values and volumes; interaction insignificant for unit values. - Alternative quality measures and samples: - Weighted and unweighted quality averages, binned quality, assigning mean ratings to unrated wines, unit-value proxies, excluding US, vintage-year disaggregation, and varying crisis window yield qualitatively similar results. - Including small volumes (<4.5 liters) slightly increases quality effect; main conclusions persist. - Mean reversion: - Including lagged dependent variables shows quality effect during crisis matters beyond mean reversion (caveat: short time dimension). - Summary of robustness tables: - Table 19 and Table 20 variants maintain negative quality × D_crisis estimates for export values and volumes across many specifications. ### X. Quantitative summaries and key magnitudes - Full-sample means and counts: - Observations: full sample 8,051; pre-crisis 4,344; crisis 3,707. - Wine Spectator score mean: 84; Parker score mean: 88. - Number of firms mean: 138 (pre-crisis 126; crisis 115). - Number of wines mean: 1,167 (pre-crisis 996; crisis 945). - Number of destinations mean: 84 (pre-crisis 81; crisis 76). - Aggregate decline and margin decomposition: - Total exports decline between 2007Q4–2008Q3 and 2008Q4–2009Q3: −29.48 percent. - Intensive margin accounts for 63.27 percent of collapse; extensive margin 36.73 percent. - Intensive components: quantities −21.99 percent; prices +4.28 percent. - Destination concentration examples: - United States export share 15.91 (2007Q4–2008Q3) and 15.83 (2008Q4–2009Q3); United Kingdom 7.28 and 6.44. ### XI. Interpretation, caveats, and implications - Principal interpretation: the global financial crisis induced a flight from quality in traded Argentinean wines; higher-quality wines suffered larger declines in exports during the crisis, driven mainly by quantities. - Macro implications (if generalizable): - Trade composition by quality affects responsiveness of trade flows to downturns; countries specialized in higher-quality goods may see larger export declines in recessions. - Crisis impacts on quantities imply large real effects of financial crises. - Distributional welfare costs: flight from quality reduces consumer welfare when consumers value variety and quality. - Caveats: - Analysis limited to Argentinean wine exports; cannot infer substitution patterns across exporters (e.g., consumers switching from French/Italian wines to Argentinean wines). - Generalizability to other industries unclear. - Suggested future research: - Investigate markups during downturns to assess whether lower inflation for higher-quality wines reflects margin compression to preserve market shares. - Examine behavior of markups across destinations given identical marginal costs. *Italic: Source — _wp1630 (excerpts and tables summarized from the provided PDF content).* ### 1. Aggregate Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19 ### 1. Aggregate Demand ### Overview - Section title: Aggregate Demand - Location: page 19 *Source: _wp1630 - 1. Aggregate Demand . . . . . . . . . . . . . . . . . . .19* ### References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ### _wp1630 - References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ### I. Introduction — research question and context - Great Trade Collapse: world trade fell by 30 percent in nominal terms, and 18 percent in real terms between 2008Q3 and 2009Q2 (World Trade Monitor, 2014). - Research objective: empirically investigate whether the global financial crisis induced heterogeneous effects on traded goods differentiated by quality (a “flight from quality”). - Theoretical motivation: - Non-homothetic preference models (Fajgelbaum, Grossman, and Helpman, 2011; Hummels and Lee, 2012) predict higher-quality goods are more income elastic and therefore disproportionately affected by negative income shocks. - Flight from quality defined as substitution towards lower quality goods in response to contractions (Burstein, Eichenbaum, and Rebelo, 2005; Jaimovich, Rebelo, and Wong, 2015). - Key empirical claim: strong evidence of a flight from quality in traded goods; up to nine percentage points difference in trade performance can be explained by the quality composition of traded goods. ### Empirical setting, data, and measurement - Focus: Argentinean wine exports, exploiting a rich customs dataset (2002–2009) with firm name, destination country, date of shipment, FOB value (US$), and volume (liters); wine identified by name, grape, type, and vintage year. - Quality measurement: experts’ ratings from Wine Spectator and Robert Parker (scores on a (50,100) scale); quality per wine computed as unweighted average across same name, grape, and type (vintage ignored for product definition); weighted averages by export shares used as robustness checks. - Product definition: aggregated by wine name, type, and grape — vintage year ignored to avoid exhaustibility bias. - Price proxy: unit values = value (US$) / volume (liters); exports are FOB. - Crisis period identification for Argentine wine: visually inspected — “crisis” defined as 2008Q4 to 2009Q3. - Growth rate construction: compare consecutive four-quarter periods: - Pre-treatment (pre-crisis): 2006Q4–2007Q3 vs 2007Q4–2008Q3. - Post-treatment (crisis): 2007Q4–2008Q3 vs 2008Q4–2009Q3. - Data frequency and filters: - Quarterly aggregation (to date episode precisely). - Shipments smaller than 4.5 liters excluded. - Only wine producers retained (wholesalers and retailers excluded). - Sample span used for main analysis: 2006Q4 to 2009Q3. ### Sample composition and descriptive statistics - Matched Wine Spectator sample: 138 firms, 1,167 wines, 437 different names, 3 types, 22 grapes. - Wine Spectator score range in sample: 68 to 96. - Matched Parker sample: 107 firms, 595 wines, 246 different names, 3 types, 18 grapes. - Parker score range observed: 74 to 98. - Coverage and observations: - Full sample: 138 exporters, 1,167 wines, 84 destination countries, 8,051 observations. - Pre-crisis sample: 126 firms, 996 wines, 81 countries, 4,344 observations. - Crisis sample: 115 firms, 945 wines, 76 countries, 3,707 observations. - Wine-level continuity: - 5,478 wines in initial span; 2,184 exported between at least two (out of three) consecutive four-quarter periods. - After matching with quality ratings: 2,247 wines total and 1,167 wines matched; these represent 59 and 50 percent of the total value of red, white, and rosé wine exported over the period. - Aggregate movements (for samples including 2,247 and 1,167 wines): - Nominal exports fell by 29.48 and 36.53 percent, respectively, during the crisis. - Real exports fell by 32.37 and 38.17 percent, respectively. - Unit values rose by 4.11 and 2.57 percent, respectively. - Firm-level participation dynamics: - Among 138 firms: export growth in both periods for 103 exporters; only before the crisis for 23 firms; only during the crisis for 12 firms. - Entry/exit firms are small in export volumes (below the 80th percentile of total real exports across firms). - Destination stability: of 84 export markets, 73 served continuously, 8 dropped out, 3 entered during the crisis. - Quality distribution and concordance: - Wine Spectator mean rating ~84; Parker mean rating ~88; mean ratings stable over time. - For the 487 wines with both ratings: Pearson correlation = 0.49; Kendall concordance = 0.44. - Most wines cluster in mid-to-high quality bins (Wine Spectator and Parker). ### Main empirical findings - Pre-crisis vs crisis differential by quality: - Before the crisis: a one unit increase on the quality scale raised export growth by 2 percentage points on average. - During the crisis: a one unit increase on the quality scale reduced export growth by 2 percentage points on average. - Quantitative contribution of quality composition to crisis fall: - If quality had increased during the crisis at the highest level observed in dataset, nominal and real exports would have dropped by 38.94 and 40.63 percent, compared to actual falls of 36.53 and 38.17 percent, respectively. - If quality had fallen to the lowest level in the sample, nominal and real exports would have fallen by 30.00 and 32.39 percent, respectively. - Difference in nominal trade performance due to quality composition during the crisis: 9 percentage points. - Difference in real trade performance due to quality composition during the crisis: 8 percentage points. - Extensive vs intensive margin: - Decomposition shows most adjustment during the crisis occurred at the intensive margin. - Regression strategy focuses on intensive margin changes (difference-in-difference on firm-product-destination triplets reporting positive exports in two consecutive periods). - Quantity vs price channel: - Collapse in nominal exports for higher quality wines driven essentially by a drop in quantities rather than a change in prices (unit values). - Heterogeneity: - Flight from quality stronger in destination countries where domestic wine production allowed substitution. - Flight from quality stronger for exports of smaller firms (more sensitive to aggregate demand and more credit constrained). - Nonlinearities: marginal effect of quality is significant for higher quality wines only, and magnitude rises with quality. - Time path: - Post-crisis extension: export growth recovered more strongly for higher quality wines, suggesting trade effects were temporary. ### Channels explored and evidence - Aggregate demand sensitivity: higher-quality exports more sensitive to changes in aggregate demand → crisis income shock more detrimental to higher-quality exports. - Domestic substitution: stronger flight from quality in markets with domestic wine production/self-sufficiency. - Inventory adjustments: transaction-level frequency changes suggest inventories matter for wine exports, but effects are not heterogeneous across quality — inventory adjustments cannot fully explain flight from quality. - Firm size and credit constraints: smaller firms exhibit stronger flight-from-quality patterns, consistent with higher demand sensitivity and financial constraints. ### Robustness and extensions - Robust to: - Different measures of quality (Wine Spectator, Parker, weighted/unweighted averages). - Alternative samples and specifications. - Potential endogeneity of quality. - Specifications in first differences and in levels that incorporate the extensive margin; these show crisis collapse in higher-quality exports characterized by both slower growth and fewer exports. - Alternative starting/ending dates (results remain robust). - Nonlinear quality effects: marginal impact concentrated among higher-quality wines. - Post-crisis recovery: evidence that higher-quality exports recovered more strongly after crisis, indicating temporary nature. ### Implications and interpretation - Macro implications (assuming generalizability beyond Argentine wine): - Composition of trade matters for responsiveness of trade flows to downturns; countries specialized in higher-quality goods may see larger export declines in recessions. - Crisis impacted quantities more than prices, implying large real effects of financial crises. - Distributional consequences: flight from quality reduces consumer welfare when consumers value variety and quality. - Positioning within literature: - Contrasts with studies that find no flight from quality (e.g., Levchenko and others, 2011) or emphasize expenditure switching to domestic goods (Bems and di Giovanni, 2015). - Adds to literature using observable quality measures (experts’ scores) rather than unit-value proxies. *Source: _wp1630 - References . . . . . . . . . . . . . . . . .* ### Section 2.4, the collapse in wine exports during the crisis essentially occurred at the intensive ### _wp1630 - Section 2.4, the collapse in wine exports during the crisis essentially occurred at the intensive margin ### Data and quality bins - Wine quality categorized using the Wine Spectator (and Parker ratings in robustness checks); due to small sample of “Great” wines, “Great” and “Outstanding” are merged in some analyses. - “Good” and “Very good” wines represent the largest share of the sample (in terms of number of observations, firms, wines, destinations, and mean export shares). “Great” and “Not recommended” wines have the smallest coverages. - Full sample for the decomposition includes 2,247 wines (quality observed). ### Descriptive statistics and crisis reversal - Pre-crisis nominal export growth by quality bin (mean): - “Great”: 43.18 percent - “Outstanding”: 19.65 percent - “Very good”: 11.02 percent - “Good”: close to zero - “Mediocre”: 3.01 percent - “Not recommended”: 6.39 percent - During the crisis nominal export changes by quality bin (mean): - “Great”: -31.88 percent - “Outstanding”: -26.45 percent - “Very good”: -9.75 percent - “Good”: -1.69 percent - “Mediocre”: -4.49 percent - “Not recommended”: -12.05 percent - Value-based market share mean log changes: - Pre-crisis growth: “Great” 49.41 percent; “Outstanding” 17.58 percent - During crisis decreases: “Great” -25.18 percent; “Outstanding” -27.42 percent - Pattern consistent with a flight from quality during the crisis; similar findings when merging “Great” and “Outstanding.” ### Decomposition of margins (2007Q4–2008Q3 to 2008Q4–2009Q3) - Decomposition: X = f × d × p × x, where f = number of exporting firms, d = mean number of destinations per firm, p = mean number of products per firm-destination, x = mean sales per firm-destination-product. Changes: ∆X = ∆f × ∆d × ∆p × ∆x, with ∆x = ∆q × ∆uv. - Aggregate change in nominal wine exports during crisis: -29.48 percent. - Extensive margin changes: - Number of exporters fell by 5.33 percent (∆f = 0.9467). - Mean number of destinations fell by 6.84 percent (∆d = 0.9316). - Mean number of wines exported by each firm to each country fell by 1.71 percent (∆p = 0.9829). - Combined extensive-margin contribution: (0.9467 × 0.9316 × 0.9829 − 1) × 100 = −13.31 percent. - Intensive margin: - Mean value of exports per firm-destination-product fell by 18.65 percent (∆x = 0.8135). - Intensive margin contribution: 63.27 percent of the collapse. - Extensive margin contribution: 36.73 percent of the collapse. - Intensive-margin components: - Mean quantities exported fell by 21.99 percent (∆q = 0.7801). - Mean unit values rose by 4.28 percent (∆uv = 1.0428). - Check: (0.9467 × 0.9316 × 0.9829 × 0.7801 × 1.0428 − 1) × 100 = −29.48 percent. ### Regression analysis: quality and the crisis (intensive margin focus) - Main estimating equations: - Equation (2): ∆lnX_ijk,t = β1 quality_k + β2 quality_k × D_crisis + D_ij,t + D_grape + D_type + D_p + D_HS + ε_ijk,t — explains intensive-margin changes (log changes require positive flows across consecutive four-quarter periods). - Equation (3): ∆lnX_ijk,t = γ quality_k × D_crisis + D_ij,t + D_k + η_ijk,t — conservative specification with product fixed effects D_k (quality drops out; γ measures differential crisis impact). - Baseline results (Panel A, Table 8): - Column (2): before the crisis a one unit increase in quality raised export growth by 0.018 (two percentage points), and during the crisis reduced it by 0.036 (four percentage points); net differential during crisis = 0.018 − 0.036 = −0.018 (two percentage points reduction from pre-crisis effect). - Results robust to interacting grape, type, province, and HS fixed effects with time dummies, weighting observations by previous-period volumes, using Parker ratings, and restricting to wines exported both before and during the crisis. - Dependent-variable breakdown: - Export volumes: pattern and magnitude similar to export values. - Unit values: interactions between quality and crisis dummy are negative but mostly insignificant — collapse driven by quantities rather than prices. ### Channels explaining the flight from quality - Hypothesized primary driver: fall in aggregate demand combined with higher income elasticity for higher quality goods. - Aggregate demand proxies used (annual 2007–2009, country coverage varies): - Real GDP per capita growth (Penn World Tables) - Household real final consumption expenditures per capita (World Bank WDI) - Unemployment rate (World Bank WDI) - Sample mean changes in these proxies during crisis: - Mean real GDP per capita growth: from 1.6 to −3.9 percent. - Mean household expenditures per capita growth: from 1.6 to −2.0 percent. - Mean unemployment rate change: from −0.1 to 1.9 percent. - Results (Panel A, Table 9): - Interaction of quality with income growth is positive and significant: the drop from 1.6 to −3.9 percent was more detrimental to higher quality exports. - Similar results when using household expenditures per capita and unemployment rate (negative interaction for unemployment implies higher-quality exports reduced more when unemployment rose). - For unit values, interactions with aggregate demand proxies are insignificant. - Including product-time fixed effects (columns 4–6) leaves the quality × aggregate-demand interactions significant for nominal and real exports. - Other channels considered: expenditure switching from imports to domestic varieties, inventory adjustments, firm size. Exchange rate interactions with quality were estimated but coefficients were insignificant (sample likely lacks sufficient time-series variation). ### Country-specific effects during the crisis (selected destinations) - Country cases: - United States: real GDP per capita growth fell from -0.8 to -5.2 percent. - United Kingdom: real GDP per capita growth fell from -0.6 to -5.9 percent. - China: real GDP per capita growth remained high (8.5 and 8.8 percent before and during the crisis). - Hong Kong: income per capita growth shrank from 1.2 to -4.2 percent; ad-valorem import duties on wine cut from 40 to zero percent in February 2008; import licences and permits removed in June 2008. - Estimating equation (3) with country-specific quality × crisis interactions (Table 10): - United States (Panel A, column 1): a one unit increase in quality reduced export growth from its pre-crisis level by 11.2 percentage points relative to other countries (~ compared to around 2.5 percentage points for other countries). - United Kingdom (Panel A, column 2): a one unit increase in quality reduced export growth from pre-crisis level by 7.3 percentage points relative to other countries. - For the US both quantities (Panel B) and prices (Panel C) of higher-quality wine exports decreased more; for the UK only quantities declined more sharply. - China (column 3): interactions for nominal and real exports are insignificant; higher-quality wines in China experienced lower inflation but no differential export-growth effect. - Hong Kong (column 4): a one unit increase in quality raised growth rates of nominal and real exports from pre-crisis levels by 7.0 and 7.8 percentage points, respectively, and reduced growth rate of unit values by 0.8 percentage points — liberalization measures likely offset negative aggregate-demand effects. - Column (5) controls simultaneously for different countries; results remain similar. *Source: _wp1630 - Section 2.4, the collapse in wine exports during the crisis essentially occurred at the intensive margin* ### 2. Expenditure Switching ### 2. Expenditure Switching ### Mechanism and hypotheses - In response to an adverse income shock, households reduce both the quantity and the quality of the goods they consume, leading to a sharper contraction in higher than in lower quality imports. - The flight from quality in traded goods is expected to be stronger in countries where households can substitute imports with domestic alternatives (e.g., because imports are more expensive or due to greater patriotism/trust in domestic goods). - Expenditure switching from imported to domestically produced wines is only possible in wine producing countries. - An alternative channel considered is substitution between wines imported from different countries, but data constraints prevent addressing this possibility. ### Descriptive evidence (summary of Table 11) - Before the crisis, mean growth in nominal and real exports to producing countries: 6.06 percent and 2.61 percent. - Before the crisis, mean growth in nominal and real exports to non-producing countries: 7.49 percent and 5.04 percent. - During the crisis, nominal and real exports fell on average to wine producing countries by 10.10 percent and 16.26 percent, respectively. - During the crisis, nominal and real exports fell on average to non-producing countries by 0.76 percent and 3.15 percent, respectively. - For major producing country groups (five, ten, and 15 largest producers): - Nominal export growth to the five largest producers fell on average by 26.45 percent. - Nominal export growth to the ten largest producers fell on average by 8.13 percent. - Nominal export growth to the 15 largest producers fell on average by 8.79 percent. - For all groups of producing countries, mean export inflation was higher during the crisis than before. ### Regression evidence: quality × crisis × producer-group interactions (summary of Table 12 findings) - Equation (3) estimated allowing the quality × crisis interaction to vary across: largest producers, other producers, and non-producers. Largest producers defined alternatively as top five, ten, and 15 producers. - Panel A, column (1) (five major producers): - Quality × crisis interaction for five major producers = -0.087. - Quality × crisis interaction for other producers = -0.027. - Quality × crisis interaction for non-producers = insignificant. - Interpretation: A one unit increase in quality reduced nominal export growth from its pre-crisis level by 8.7 percentage points to the five largest producing nations, versus 2.7 percentage points to the other producers. - Columns (2) and (3) (ten and 15 largest producers): similar pattern with the magnitude for the largest producers decreasing gradually from column (1) to (3). - Similar observations apply for export volumes (Panel B) and unit values (Panel C). - Conclusion: Despite becoming cheaper in relative terms, higher quality wines suffered more dramatically when exported to the largest producing countries. - Robustness: Columns (4) to (6) include real GDP per capita growth interacted with quality; coefficients on quality × crisis become smaller and less significant but results still hold. ### Regression evidence: quality × crisis × self-sufficiency interactions (summary of Table 13 findings) - Self-sufficiency = domestic wine production / consumption (zero for non-producers; >1 for self-sufficient; <1 for non-self-sufficient). - Expectation: flight from quality more acute in self-sufficient countries where domestic supply can satisfy demand. - Table 13 (column references correspond to export values, volumes, unit values): - For export values (column 1): - Quality × crisis interaction is insignificant for non-producing nations. - Quality × crisis interaction = -0.026 for non-self-sufficient countries. - Quality × crisis interaction = -0.073 for self-sufficient destinations. - For export volumes and unit values (columns 2 and 3): interaction is significant for self-sufficient countries only. - Columns (4) to (6) control for income per capita growth interacted with quality; qualitative results remain similar. ### Additional notes on producer group composition - With the exception of China, the five major producing countries are France, Italy, Spain, and the United States; these countries were all severely hit by the crisis. - Correlation between the five largest wine producing nations and the self-sufficient countries = 0.49. - An exploration using the 2005–2006 World Values Survey on patriotism (share “Very proud” or “Quite proud”) interacted with quality produced an insignificant coefficient (sample reduced to 36 countries due to coverage). ### Findings synthesis on expenditure switching - Empirical evidence indicates a stronger flight from quality during the crisis when destination countries are wine producing and, more sharply, when they are self-sufficient. - The largest producing destinations saw the steepest declines in higher quality exports (e.g., -0.087 interaction for five largest producers), consistent with expenditure switching toward domestically produced wines. ### Inventory Adjustments (Section 3 summary and relation to expenditure switching) - Inventory adjustments are a plausible amplifier of trade collapses: importers may deplete inventories instead of importing more in response to negative demand shocks. - Method: use change in the frequency of shipments at the transaction level as an indicator of inventory depletion (lower frequency → possible depletion). - Regression extension of equation (3) includes log change in frequency of shipments, on its own and interacted with quality (product-time fixed effects controlled). - Table 14 findings (columns (1)–(3)): - Change in frequency is positive and significant (as expected). - Interaction of frequency change with quality is: - Insignificant for export values and volumes. - Negative for unit values. - Wine type heterogeneity (columns (4)–(6)): - White and rosé wines need to be sold within a few months; red wines can be stored. - Frequency is positive and significant for red wines only. - Interaction of frequency with quality remains insignificant (negative for unit values). - Conclusion: Inventory adjustments matter for explaining trade flows generally, but their effect is not heterogeneous across goods differentiated by quality and thus cannot explain the observed flight from quality during the crisis. *Source: _wp1630 - 2. Expenditure Switching* ### 4. Firm Size ### 4. Firm Size ### Hypotheses and mechanisms - Flight from quality predicted to be stronger for exports of smaller firms because: - Importers may cut purchases from small wineries and continue to import from largest producers who are more important business partners. - Small firms are typically more restricted in their access to external finance than large firms, so tightening credit conditions during the crisis could reduce small-firm exports more (Bricongne and others, 2012; Chor and Manova, 2012). - Higher quality wine exports may be more vulnerable to tighter credit because they face higher (per unit) transportation and distribution costs which are incurred before export revenues are realized. - Domestic credit conditions in Argentina became tighter during the crisis: the daily interbank lending rate rose from an average of 8.8 to 10.9 percent between the 2007Q4–2008Q3 and 2008Q4–2009Q3 periods (data from Chor and Manova, 2012). ### Measurement of firm size and empirical approach - Firms ranked by total volume of exports (in liters). - Small group: smallest 80 percent of exporters (based on the 80th percentile). - Large group: remaining top 20 percent. - Estimated equation (3) with quality interacted with the crisis dummy varying by firm size. - Results reported across columns referenced (Table 15, columns 1–6). ### Key findings on firm size - In both nominal and real terms, exports of higher quality wines collapsed to a larger extent during the crisis for small firms; no significant effect for large firms (columns 1 and 2 of Table 15). - Results for unit values are insignificant (column 3). - Including income growth interacted with quality by firm size (columns 4–6): - For nominal exports (column 4), the coefficient for small firms is reduced in magnitude but remains significant. - Interpretation: demand-side factors do not fully explain the collapse of small firms’ higher quality exports; supply-side factors such as tighter credit conditions may have played a role (unit value results remain insignificant). ### Extensions discussed in this unit - The section examines nonlinearities in the effect of quality, extensive margin adjustments, and the post-crisis recovery. --- ### Nonlinearities - Estimated equation (2) including quality and quality squared, on their own and interacted with the crisis dummy. - Nonlinearities present for export values, volumes, and (less so) unit values (Table 16, columns 1–3). - Negative squared quality terms during the crisis indicate semi-elasticities of export values and volumes to quality increase in magnitude at higher levels of quality. - Semi-elasticities evaluated at mid scores of quality bins (Table 1); evaluation points: 62 for “Not recommended,” 77 for “Mediocre,” 82 for “Good,” 87 for “Very good,” 92 for “Outstanding,” and 97.5 for “Great” wines. - Pre-crisis semi-elasticities (column 1) for nominal export growth: - Positive and significant for “Good,” “Very good,” “Outstanding,” and “Great” wines only. - Magnitude rises with quality: from 0.015 for “Good” to 0.061 for “Great” wines. - During the crisis: - Semi-elasticity negative and significant for higher quality wines only. - Magnitude increases with quality: from -0.014 for “Good” to -0.072 for “Great” wines. - Positive coefficient for “Not recommended” wines indicates their exports have grown. - Columns (2) and (3): before the crisis stronger growth for higher quality wines driven by quantities rather than prices. - During the crisis lower growth for “Very good,” “Outstanding,” and “Great” wines driven by lower growth in both quantities and prices (and lower inflation only for “Good” wines). - Robustness: Columns (4) to (6) include product fixed effects; results remain consistent (prices semi-elasticities insignificant). - Using column estimates, Figure 4 plots semi-elasticities across quality scores (68 to 96): - Before the crisis: semi-elasticities for export values and volumes positive and significant for quality levels above 81; unit values insignificant. - During the crisis: semi-elasticities positive at low quality levels (between 68 and 71), become negative for all wines above 81 for export values and above 86 for volumes. - For unit values during the crisis: semi-elasticities negative and significant for wines above 79. --- ### Extensive margin - First-difference specifications handle intensive margin; extensive margin also materially relevant. - To incorporate extensive margin: - Dependent variable: mid-point export growth rate g_i jk,t = (X_i jk,t − X_i jk,t−1) / [1/2 (X_i jk,t + X_i jk,t−1)] where X_i jk,t can be zero; this measure is symmetric and bounded between -2 and +2. - Mid-point growth rate equals +2 for an entry, -2 for an exit, values between zero and +2 if flows increased, and between -2 and zero if decreased. - Alternative dependent variable: first difference of ln(1+X_i jk,t). - Level specifications: ln X_i jk,t (excludes zero trade observations). - Tobit regression of log(exports) censored at minimum observed positive value by destination (Crozet and others, 2012). - Poisson Pseudo-Maximum Likelihood (Santos Silva and Tenreyro, 2006). - Heckman selection correction not implemented due to unavailable exclusion restriction. - Results (Table 17): - Despite variation in magnitudes across specifications, patterns broadly support baseline results (interaction insignificant for export volumes in column 3). - Level specifications (columns 3–5) indicate the collapse in higher quality exports characterized by both slower growth and lower export levels. - Interactions between quality and crisis dummy are smaller in columns (4) and (5) once extensive margin accounted for. --- ### Post-crisis recovery - Objective: test whether higher quality exports recovered more strongly after the world economy recovery; investigate temporary vs. persistent effects. - Data limitation: 2010 customs dataset records wine names as missing (firm names, destinations, grapes, types, vintage years available), so wines exported in 2010 cannot be identified. - Consequently, to compare identical quarters and avoid seasonality, earliest post-crisis usable period is 2011Q4–2012Q3. - Growth for post-crisis measured between 2008Q4–2009Q3 and 2011Q4–2012Q3 (two-year gap; growth rate over a three-year period given missing 2010). - Full sample spans three periods: pre-crisis, crisis, post-crisis. - Estimated equation (5): ∆lnX_i jk,t = α1 quality_k + α2 quality_k × D_crisis + α3 quality_k × D_post + controls + ε_i jk,t - D_post = 1 for 2011Q4–2012Q3 post-crisis period. - α1 captures pre-crisis effect of quality; α2 and α3 measure differential effects during and after crisis. - Expected signs from earlier results: α1 > 0, α2 < 0, α1 + α2 < 0. - Recovery interpretation: quality effect during recovery = α1 + α3. If higher quality exports recovered more strongly, α3 should be either insignificant, positive, or negative but with |α1| > |α3| so α1 + α3 > 0. - Results (Panel A, Table 18, column 1): - Export growth stronger for higher quality wines both before and after the crisis, and fell more during the downturn. - A one unit increase in quality strengthened export growth by about two percentage points before the crisis, and subsequently lowered it by two percentage points. - Positive coefficient on quality × post-crisis indicates one unit increase in quality raised export growth by 6.3 percentage points(0.018+0.045) over the period. - Caveat: post-crisis growth calculated over a three-year period due to unavailable 2010 data. - Robustness variants: - Column (2) excludes post-crisis interaction: higher quality exports increased more in “normal” periods (before and after crisis combined) but fell more during the crisis. - Columns (3) and (4) exclude pre-crisis and crisis data respectively; results consistent. - Columns (5)–(8) include product fixed effects; crisis and post-crisis interactions remain negative and positive, respectively. - Panels B and C: - Export volumes results comparable to values both quantitatively and qualitatively. - Some indication that inflation fell more for higher quality wines during the crisis, though not always significant. - Inflation of higher quality wine exports stronger after the downturn. - Conclusion on recovery: - Crisis had heterogeneous effects by quality. - Flight from quality in traded goods was temporary; recovery in higher quality exports occurred after the crisis period. --- *Source: _wp1630 - 4. Firm Size (extracts from the provided PDF content).* ### 38.94 percent, compared to an actual fall of 36.53 percent. In other words, quality upgrading ### _wp1630 - 38.94 percent, compared to an actual fall of 36.53 percent. In other words, quality upgrading ### Main findings on quality effects during the crisis - Nominal export growth: quality upgrading during the crisis would have reduced nominal export growth by 2.4 additional percentage points (comparison: predicted fall of 38.94 percent, actual fall of 36.53 percent). - Alternative extreme scenario (quality scores fall to lowest sample value of 68): predicts exports would have dropped by 30.00 percent. - Difference between the two opposite scenarios (quality upgraded to 96 vs downgraded to 68) implies a difference in export performance of about nine percentage points. - Real export quantities: actual decrease in export volumes was 38.17 percent. - With quality upgrading (quality = 96): export quantities would have decreased by 40.63 percent. - With quality downgrading (quality = 68): export quantities would have decreased by 32.39 percent. - The difference in export growth predicted by the two opposite scenarios is about eight percentage points. ### Robustness checks and alternative specifications - Endogeneity - Instrumental-variable approach: Parker scores used as an instrument for Wine Spectator ratings (both interacted with the crisis dummy) under the assumption of uncorrelated measurement errors. - Results: instrumented coefficients increase in magnitude for both export values and volumes; interaction is insignificant for unit values. - Kleibergen-Paap F statistic = 437 (critical value = 16, Stock and Yogo, 2005), rejecting weak instrument null. - First-stage estimated coefficient for Parker on Wine Spectator (interacted with crisis) = 0.703, significant at the one percent level. - Quality measurement alternatives - Weighted average of quality scores (weights = share of exports of each wine in total exports of wines with same name, grape, and type, by destination and time); correlation between weighted and unweighted quality = 89 percent (results in column (2) of Table 19). - Binned quality variable: values 1–6 corresponding to Wine Spectator bins (Table 1) used to minimize noise (column (3)). - Mean Wine Spectator rating by wine name and type assigned to unrated wines to increase coverage (column (4)). - Dropping the US from the sample to address US-centered ratings concern (column (5)). - Unit-value proxies for quality: - Log mean unit value of exports across destinations and over time. - Product ranks by mean unit values per exporter and time period (highest unit value rank = 1, etc.) used as inverted indicator of quality (columns (6) and (7)). - Overall: in most cases, results continue to hold and are in line with expectations. - Sampling and crisis-period variations - Extend pre-crisis sample up to 2005Q4 (adds pre-treatment growth between 2005Q4–2006Q3 and 2006Q4–2007Q3) (column (1) of Table 20). - Extend pre-crisis sample up to 2003Q4 (excluding 2002 due to Argentine recession) so pre-crisis sample includes growth rates over four periods (column (2)). - Vary crisis episode length: - Drop 2008Q4 from crisis period; growth rates measured over three-quarter periods (column (3)). - Include 2009Q4 in the crisis period (column (4)). - Vintage year disaggregation - Define product including vintage year (sample: 99 firms, 1,025 wines, 73 destination countries); original Wine Spectator scores used though sample more unbalanced and reduced by more than half. - Results: interactions between quality and crisis indicator larger in magnitude for nominal and real exports compared to baseline (unit values insignificant), but qualitatively similar (column (5) of Table 20). - Mean reversion - Higher-quality wines grew most before the crisis and had largest drops during the recession; include lagged dependent variables to control for mean reversion. - Results (column (6) of Table 20) show quality effect during the crisis matters beyond mean reversion. - Note: short time dimension prevents use of dynamic panel data estimation techniques for within-group lagged dependent variables. - Aggregation across destinations - Aggregate exports of each product by firm across destinations; replace firm-destination-time fixed effects with firm-time dummies and cluster standard errors by firm-time (column (7) of Table 20). - Small volumes - Including volumes less than 4.5 liters (column (8) of Table 20) yields slightly larger quality effect; overall results remain robust. ### Interpretation, caveats, and avenues for future research - Principal interpretation: the global financial crisis induced a flight from quality in traded Argentinean wines, with higher-quality wines experiencing larger declines in exports during the crisis. - Caveats: - Analysis limited to Argentinean wine exports; cannot determine if consumers in crisis-hit countries substituted from more expensive French or Italian wines to cheaper Argentinean wines. - Unclear if results generalize to other industries. - Suggested future research: - Investigate behavior of markups during the downturn to assess whether decreases in inflation for higher-quality wines resulted from firms compressing margins to preserve or restore export market shares. - Note: price differences for the same wines across destinations reflect differences in markups given identical marginal cost. *Source: Excerpt from IMF working paper content provided in the source PDF.* ### 09. IMF Economic Review 60(3), 303–328. ### _wp1630 - 09. IMF Economic Review 60(3), 303–328. ### Data and sample overview - Full sample observations: 8,051. - Pre-crisis observations: 4,344. - Crisis observations: 3,707. - Variables (means, full sample unless noted): - Export values growth (%) — Mean: 0.24; Pre-crisis mean: 6.55; Crisis mean: -7.15. - Export volumes growth (%) — Mean: -3.73; Pre-crisis mean: 3.43; Crisis mean: -12.12. - Unit values growth (%) — Mean: 3.97; Pre-crisis mean: 3.12; Crisis mean: 4.97. - Wine Spectator score — Mean: 84 (full sample). - Parker score — Mean: 88 (sample where available). - Number of firms — Mean: 138 (full sample); Pre-crisis: 126; Crisis: 115. - Number of wines — Mean: 1,167 (full); Pre-crisis: 996; Crisis: 945. - Number of destinations — Mean: 84 (full); Pre-crisis: 81; Crisis: 76. ### Quality classification and basic patterns - Wine Spectator quality bins and mean unit values (US$ per liter) and mean export shares (%): - “Outstanding” and “Great”: Mean unit value 14.60; Pre-crisis 14.17; Crisis 15.16. Mean export share full sample 9.58; Pre-crisis 10.43; Crisis 8.32. - “Great”: Mean unit value 22.38; Pre-crisis 27.17; Crisis 20.47. Mean export share full sample 0.02; Pre-crisis 0.01; Crisis 0.03. - “Very good”: Mean unit value 4.57; Pre-crisis 4.58; Crisis 4.55. Mean export share full sample 39.86; Pre-crisis 41.50; Crisis 37.47. - “Good”: Mean unit value 4.24; Pre-crisis 4.15; Crisis 4.35. Mean export share full sample 42.56; Pre-crisis 39.81; Crisis 46.55. - “Mediocre”: Mean unit value 3.49; Pre-crisis 3.06; Crisis 3.99. Mean export share full sample 5.90; Pre-crisis 6.34; Crisis 5.27. - “Not recommended”: Mean unit value 2.64; Pre-crisis 2.59; Crisis 2.71. Mean export share full sample 2.10; Pre-crisis 1.91; Crisis 2.39. ### Main empirical finding: Quality and the crisis - Baseline effect of quality interacted with crisis on exports (select estimates, Table 8): - Panel A (Export values): quality × D_crisis = −0.036 (0.013) [column (1)], significant at 1% level. - Panel B (Export volumes): quality × D_crisis = −0.031 (0.013) [column (1)], significant at 5% level. - Panel C (Unit values): quality × D_crisis = −0.005 (0.003) [column (1)], weakly significant at 10% level. - Interpretation: Higher-rated wines experienced larger declines in export values and volumes during the crisis period; unit-value declines are smaller and less consistently significant. ### Aggregate demand and quality interaction - Interaction of aggregate demand changes with quality (Table 9, Panel A and B): - ∆lnAD × quality on export values: 0.411 (0.139) [column (1)], significant at 1% level. - ∆lnAD × quality on export volumes: 0.397 (0.133) [column (1)], significant at 1% level. - Panel C (Unit values): ∆lnAD × quality estimates not significant in columns shown. - Interpretation: Destination aggregate demand shifts amplify the sensitivity of exports and volumes to quality. ### Heterogeneity by destination and market type - Destination-specific crisis interactions (Table 10, Panel A & B): - quality × D_crisis × US on export values: −0.112 (0.023) [column (2)], significant at 1% level. - quality × D_crisis × UK on export values: −0.073 (0.023) [column (3)], significant at 1% level. - quality × D_crisis × China on unit values: −0.027 (0.012) [column (4)], significant at 5% level. - quality × D_crisis × Hong Kong on unit values: −0.008 (0.003) [column (5)], significant at 1% level. - Wine producing vs non-producing and largest markets (Table 12, Panel A–C): - quality × D_crisis × largest (big wine-producing destinations) on export values: −0.087 (0.021) [column (1)], significant at 1% level. - quality × D_crisis × non-producers on export values: not significant in many specifications. - Unit values show stronger negative quality × crisis effects for largest destinations: e.g., −0.013 (0.004) [Panel C, column (1)], significant at 1% level. - Interpretation: The adverse crisis impact on higher-quality wines is concentrated in large and certain developed destination markets, notably the US and UK. ### Margins and decomposition - Extensive vs intensive margin decomposition (Table 7): - Total exports (2007Q4–2008Q3): 178,071 (thousand US dollars); 2008Q4–2009Q3: 125,585. - Growth: −29.48%. - Contributions: Extensive margin contribution 36.73%; Intensive margin contribution 63.27%. - Intensive margin components growth: sales −18.65%; quantities −21.99%; prices +4.28%. - Extensive margin regressions (Table 17): - quality × D_crisis effect on export values (Panel A, column (1)): −0.024 (0.009), significant at 5% level. - Various specifications show negative quality × crisis effects on the extensive margin across estimators (OLS, Tobit, Poisson). ### Inventory adjustments and product characteristics - Inventory/frequency effects (Table 14): - ∆ln frequency on export values: 1.589 (0.781) [column (1)], significant at 5% level. - ∆ln frequency × quality on unit values: −0.004 (0.002) [column (3)], significant at 5% level. - Red wine interaction: ∆ln frequency × red on export values 1.742 (0.781) [column (3)], significant at 5% level. - Interpretation: Firms adjusted export frequency during the crisis; frequency shifts are associated with changes in export values and unit values, with some heterogeneity by wine type (red vs white/rosé). ### Firm size heterogeneity - Quality × D_crisis × firm size (Table 15): - Interaction for small firms on export values: −0.048 (0.019) [column (1)], significant at 1% level. - Interaction for large firms: −0.018 (0.016) [column (1)], not significant. - Interpretation: Small firms' higher-quality products faced larger crisis-related declines than large firms. ### Nonlinearities and semi-elasticities by quality bin - Nonlinear specification (Table 16) and semi-elasticities (selected): - Pre-crisis semi-elasticities (export values): - “Great”: 0.061 (0.031), significant at 5% level. - “Outstanding”: 0.045 (0.019), significant at 5% level. - “Very good”: 0.030 (0.010), significant at 1% level. - Crisis semi-elasticities (export values): - “Great”: −0.072 (0.027), significant at 1% level. - “Outstanding”: −0.051 (0.030), significant at 1% level. - “Very good”: −0.033 (0.013), significant at 5% level. - Unit values semi-elasticities show negative effects during crisis for higher-quality bins (e.g., “Great” unit values: −0.022 (0.007), significant at 1%). - Interpretation: Relationship between quality and export performance reverses from positive pre-crisis to negative during the crisis, particularly for top-quality bins. ### Post-crisis recovery patterns - Quality interactions across pre-crisis, crisis, and post-crisis periods (Table 18): - quality × D_crisis on export values: −0.036 (0.013) [column (1)], significant at 1% level. - quality × D_post on export values: 0.045 (0.017) [column (1)], significant at 1% level. - Similar signs for export volumes: quality × D_post = 0.029 (0.014), significant at 5% [Table 18, Panel B]. - Unit values: quality × D_post = 0.016 (0.007), significant at 5% [Table 18, Panel C]. - Interpretation: Higher-quality wines showed signs of recovery post-crisis, with positive interactions in the post period offsetting crisis-period negative effects. ### Endogeneity checks and robustness - Endogeneity and robustness on quality (Table 19): - IV and alternative estimators produce negative quality × D_crisis estimates on export values and volumes; e.g., quality × D_crisis = −0.072 (0.032) [Panel A, column (1)], significant at 5% for IV. - Weighted and alternative subsamples maintain qualitatively similar negative crisis interactions. - Other robustness checks (Table 20): - Inclusion of lagged dependent variables, different sample windows, vintages, and minimum destination volume thresholds produce consistent negative quality × D_crisis estimates for export values and volumes (e.g., −0.032 (0.011) [Panel A, column (1)], significant at 1%). - Unit values retain negative or small effects under many robustness checks; e.g., −0.007 (0.002) [Panel C, column (1)], significant at 1%. ### Key quantitative summaries and magnitudes - Aggregate loss in total exports between 2007Q4–2008Q3 and 2008Q4–2009Q3: from 178,071 to 125,585 (thousand US dollars), a growth (decline) of −29.48%. - Intensive margin decomposition: quantities down −21.99%; prices up 4.28% over the same periods. - Destination concentration (Table 2): United States export share 15.91 (2007Q4–2008Q3) and 15.83 (2008Q4–2009Q3); Netherlands 7.34 and 7.42; United Kingdom 7.28 and 6.44. *Italic: Source — _wp1630 - 09. IMF Economic Review 60(3), 303–328.* --- _Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1630.pdf_