University of Sussex, PhD Conference 5th of December 2014 Catalysts and Inhibitors of the Trade Collapse Mattia Di Ubaldo Motivation & Context - 1 The Slovenian Experience Growth of trade Growth of GDP 29/10/2014 Mattia Di Ubaldo, University of Sussex 2 Motivation & Context - 2 • Causes of trade collapse: • Supply side • Credit crunch – working capital (Bricongne et al., JIE 2012; Chor & Manova, JIE 2012; Behrens et al., RES 2013). • Trade finance – bank or firm intermediated (Korinek et al., 2010; Malouche, 2011; Antràs & Foley, JPE 2014). • Demand side • Decrease in expenditure – composition (Engel & Wang, JIE 2009; Eaton et al., 2011; Petropoulou & Soo, 2011). • Vertical linkages (Levchenko et al., IMF review 2010, Bems et al., AER 2011). • Inventory adjustments (Alessandria et al., AER 2010). 29/10/2014 Mattia Di Ubaldo, University of Sussex 3 Contribution My analysis: factors that amplified or dampened the reaction of trade. • Most of the action was in GVCs! Intermediates: • 2/3 of total trade (Bems et al., 2011). • 11/15 most affected sectors (Bricongne et al. 2012). 1. Analyse the reaction of different inputs, depending on the cost-share in firms’ sales. 2. Explore the reaction of inputs along Intra-Firm vs Arm’s Length trade dimension: one WP so far (Altomonte et al. 2012), but I add: • Cost-share of inputs. • Intensive and Extensive margins. • Firm fixed effects and additional firm controls. 3. Detailed intensive/extensive (firm, destination, product) margins decomposition. • New across Intra-firm vs Arm’s Length trade. 29/10/2014 Mattia Di Ubaldo, University of Sussex 4 Preview of findings 1. Cost-share of imported inputs in sales: • Higher cost share, larger reaction (but Intra-Firm dampens). 2. Firm affiliation – Intra-firm (RP) versus arm’s length (AL) trade. • No different performance. 3. Intensive vs Extensive margin • 29/10/2014 70% Intensive margin; larger intensive margin share for RP. Mattia Di Ubaldo, University of Sussex 5 Outline • Data and sample. • Hypotheses. • Methodology. • Results. 29/10/2014 Mattia Di Ubaldo, University of Sussex 6 Data and Sample 29/10/2014 Mattia Di Ubaldo, University of Sussex 7 Data • Slovenia: why? • small, open economy. • strongly integrated with both eastern and western European countries: intermediates 72% of imports. • Transaction-level trade data (SURS): monthly exports/imports at the CN-8 digits product level (xickt); 2000-2011. • Firm Balance Sheet data (AJPES): balance sheet and income statements of all Slovenian firms; 2000-2011. • Ownership data (Bureau Van Dijk): ORBIS allows to track all the proprietary network, on a world scale, up to the 10th level of subsidiarity. 29/10/2014 Mattia Di Ubaldo, University of Sussex 8 Intra-firm (RP) trade proxy • I need to infer whether shipments are RP or AL: • I assume that transactions are RP when there is an affiliate in the destination they are directed to. 88% of trade by affiliates to a certain destination is either pure RP or pure AL (Bas & Carluccio 2009), • This inflates RP trade proxy when firms adopt a mixed strategy; but it’s a classification mistake working against me. • 49% exports are RP in Slovenia in 2007: • 47% in US (US Census Bureau data) • 52.6% exports to US are RP – Slovenian data; 51.3% – Census Bureau data (Lanz & Miroudot, 2011). 29/10/2014 Mattia Di Ubaldo, University of Sussex 9 Timing and Sample Timing October 2008 – May 2010; trough in November 2009. Final sample 29/10/2014 Exporters 9,238 NACE-4dig sectors 462 Products 7,350 median 58; mean 136 Destinations 199 median 10; mean 16 Mattia Di Ubaldo, University of Sussex 10 Hypotheses 29/10/2014 Mattia Di Ubaldo, University of Sussex 11 Hypotheses – 1 and 2 • Inventory adjustments were observed to have amplified the reaction of trade (Alessandria et al. 2010). GVC ideal locus! 1st Hypothesis: a higher cost−share of imported intermediates in sales leads to a larger reaction of trade. Why? Inventories of higher cost share inputs get adjusted more promptly, if inventories management costs are proportional to the cost share of inputs. Larger adjustments! 2nd Hypothesis: RP trade of intermediates is more resilient than AL trade. Why? For RP trade: • Lower uncertainty • Shorter delivery lags • Better communication 29/10/2014 Lower buffer of inventories Mattia Di Ubaldo, University of Sussex Lower adjustments! 12 Hypotheses – 3 and 4. 3rd Hypothesis: intermediates’ inventories adjustment generate a Bullwhip Effect: this is exacerbated by the cost share in sales. Why? Inventories accelerator mechanism! • The reaction of RP and AL trade can differ across trade margins: In a crisis, different sunk costs, market rigidities and hold-up problem would cause: 4th Hypothesis: intensive margin adjustments are more pronounced for RP trade than for AL trade; vice-versa for extensive margin adjustments. Why? Firms might more easily reduce the size of shipments to affiliates rather than to non-affiliates. Extensive margin changes could happen more at AL. 29/10/2014 Mattia Di Ubaldo, University of Sussex 13 Methodology 29/10/2014 Mattia Di Ubaldo, University of Sussex 14 Methodology Hypothesis 3 g ickt α 0 α1Ω α 2Ω* recovery γ i ε ickt Dep. var.: g ickt x ickt x ick(t 12) 0.5(x ickt x ick(t 12) ) Ω β1Int ickt β 2 RPikt β3CS k δ1 Int ickt * CS k δ2 Int ickt * RPikt δ 3 Int ickt * CS k * RPikt β4 X iy β5 ExYug ickt Hypothesis 1 Hypothesis 2 im ickt 1 t 1 CS k Y NY y 2000 i 1 iy 2007 29/10/2014 N 12 Mattia Di Ubaldo, University of Sussex 15 Results Estimations 29/10/2014 Mattia Di Ubaldo, University of Sussex 16 RP trade and Cost Share (Hyp. 1 and 2): (1) Int. (4) (5) (6) (7) 0.0389*** 0.0389*** 0.0559*** 0.0588*** 0.0612*** (3.04) (3.79) (4.62) (4.78) (4.97) -0.0448 -0.0465 -0.0482 -0.0480 -0.0616* (-1.39) (-1.25) (-1.26) (-1.26) (-1.64) -0.00391 -0.00156 -0.000922 -0.0104 (-0.70) (-0.29) (-0.17) (-0.27) 0.00166 0.000918 -0.00934 -0.0104 (0.05) (0.02) (-0.24) (-0.27) -0.178 -0.287* -0.306** (-1.26) (-1.95) (-2.10) 0.380* 0.356 (1.62) (1.54) RP (2) CS (3) Int. * RP Int. * CS Int. * CS * RP -0.106*** Ex Yug. (-4.05) Firm FE yes yes yes yes yes yes yes Firm Controls yes yes yes yes yes yes yes N 1,636,936 1,636,936 1,636,936 1,636,936 1,636,936 1,636,936 1,636,936 Note: t statistics arising from robust standard errors clustered at the firm level in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01. Results: sector-product CS variable • Sectoral disaggregation of the product level CS variable: (1) im ijckt 1 t 1 CS k j Y NY y 2000 i 1 ijy N (5.31) (5.48) -0.0468 -0.0612 (-1.16) (-1.55) 0.000588 0.000871 0.000858 (0.74) (1.12) (1.09) -0.00944 -0.0108 (-0.23) (-0.26) -0.125** -0.143** (-2.20) (-2.27) RP CS-SECT 2007 0.0607*** (3) 0.0633*** Int. 12 Int. * RP Int. * CS-SECT (2) 0.000236 Int. * CS-SECT * RP (0.00) 29/10/2014 Firm FE yes yes yes Firm Controls yes yes yes N 1,554,771 1,554,771 1,554,771 Note: t statistics arising Mattia Di Ubaldo, University of from Sussexrobust standard errors clustered at the firm level in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01. 18 Hypothesis 3: Bullwhip (1) Int. Downturn (2) (3) 0.0113 0.0234 (0.72) (1.44) CS -0.00855 -0.00636 (-1.20) (-0.90) -0.294** Int. * CS (-2.05) Int. * Rec Recovery 0.0958*** 0.0864*** (3.59) (3.24) CS * Rec. 0.0157 0.0167 (1.09) (1.16) 0.401** Int. * CS * Rec. (2.31) Firm FE yes yes yes Firm Controls yes yes yes N 1,636,936 1,636,936 1,636,936 Note: t statistics arising from robust standard errors clustered at the firm level in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01. Results Margin decomposition 29/10/2014 Mattia Di Ubaldo, University of Sussex 20 Margin decomposition -2 Decomposition of growth of exports: net intensive and net extensive margin contributions 29/10/2014 Mattia Di Ubaldo, University of Sussex 21 Intensive/Extensive margin decomposition (in %): RP vs AL trade 29/10/2014 Mattia Di Ubaldo, University of Sussex 22 Conclusion • This work adds to the study of the trade collapse by finding: • RP trade reacted more at the intensive margin, compared to AL trade; but no significant difference overall. • A higher cost-share of inputs induced a larger reaction of trade. • Especially larger fall in downturn. • CS heterogeneity allows to discover: • Dampening impact of RP trade, relative to AL trade. • Bullwhip effect. 29/10/2014 Mattia Di Ubaldo, University of Sussex 23 THANK YOU! 29/10/2014 Mattia Di Ubaldo, University of Sussex 24 RP trade and Cost Share (Hyp. 1 and 2): • Impact of Cost Share of inputs and RP trade on growth rate of export of intermediates. 29/10/2014 Mattia Di Ubaldo, University of Sussex 25 Exports over the crisis Growth of Exports, with “contributions” of RP and AL trade to overall variation. 29/10/2014 Mattia Di Ubaldo, University of Sussex 26 Hypotheses – 5 and 6. Trade finance. • Trade finance (bank intermediated) became more expensive in the crisis: 5th Hypothesis: greater reliance on firm intermediated trade finance induced a better trade performance. 6th Hypothesis: greater reliance on firm intermediated trade finance induced a better trade performance for RP trade, relative to AL trade. Why? Firm intermediated trade credit should be easier between related parties. 29/10/2014 Mattia Di Ubaldo, University of Sussex 27 Methodology g ickt α 0 α1Ω γ i ε ickt Dep. var.: g ickt x ickt x ick(t 12) 0.5(x ickt x ick(t 12) ) Ω β1Int ickt β 2 RPickt δ5 TC iy δ6 TC iy * RPickt β4 X iy β5 ExYug ickt Hypothesis 5 Hypothesis 6 TC iy 29/10/2014 receivable s y 1 sales y 1 1 receivables ijy TC j n 1 y 2000 NY sales ijy N Mattia Di Ubaldo, University of Sussex 2007 28 Results – hypothesis 4 and 5: trade finance (1) (2) TC-SECT 0 0 - RP -0.0670** TC-FIRM (3) (4) (5) 0.0461* 0.0347 0.0350 (1.62) (1.38) (1.36) -0.0843* -0.0618 (-1.78) (-0.77) 0.162 0.0827 (1.40) (0.36) RP (-2.37) TC-SECT * RP 0.00312*** 0.00318*** (7.41) (9.65) -0.000752 Int. * TC-SECT * RP TC-FIRM * RP 0.119 Int. * TC-FIRM* RP (-1.11) (0.56) Firm FE yes yes Firm FE yes yes yes Firm Controls yes yes Firm Controls yes yes yes N 1,661,136 1,661,136 N 1,661,075 1,661,075 1,661,075 Note: t statistics arising from robust standard errors clustered at the firm level in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01. 29/10/2014 Mattia Di Ubaldo, University of Sussex 29 Results – hypothesis 4 and 5: trade finance • Impact of Receivables and RP trade on growth rate of exports. Larger reliance on trade credit was associated to a better trade performance, with an additional premium detected to RP trade. 29/10/2014 Mattia Di Ubaldo, University of Sussex 30 Results – all hypotheses together Int. (1) (3) 0.0575*** Int. (4.62) RP -0.0786*** (3.94) RP (-2.61) CS -0.00106 0 CS -0.287** TC-FIRM 0.381* Int. * CS 0.00330*** Int. * CS * RP -0.000947 0.386* (1.65) TC-FIRM * RP (9.31) Int.*TC-SECT*RP -0.287** (-1.96) (1.63) TC-SECT * RP 0.0321 (1.18) (-1.95) Int. * CS * RP -0.000265 (-0.05) (.) Int. * CS -0.0465 (-1.25) (-0.20) TC-SECT (4) 0.0620*** 0.0145 (0.06) Int.*TC-FIRM*RP (-1.38) 0.192 (0.81) Firm FE yes Firm FE yes Firm Controls yes Firm Controls N 1,636,936 N yes 1,636,875 Note: t statistics arising from robust standard errors clustered at the firm level in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01. Margin decomposition -1 • Methodology (Bricongne et al., 2012). I decompose mid-point growth rates: g ickt x ickt x ick(t 12) 0.5(x ickt x ick(t 12) ) Extensive margin: created (gickt =2) and destroyed flows (gickt =-2) Intensive margin: increased (0< gickt <2) and decreased flows (-2< gickt <0) Weight each flow by its share in total Slovenian exports: s ickt x ickt x ick(t 12) x i c k ickt i c k x ick(t 12) Aggregate subsets of the weighted mid-point growth rates to obtain the margins, since the aggregate year on year total growth rate can be correctly approximated by: G t c i k g ickt * s ickt 29/10/2014 Mattia Di Ubaldo, University of Sussex 32 Margin decomposition -3 Decomposition of growth of exports: detailed extensive margin contributions 29/10/2014 Mattia Di Ubaldo, University of Sussex 33 Margin decomposition -5 29/10/2014 Mattia Di Ubaldo, University of Sussex 34 Margin decomposition -4 29/10/2014 Mattia Di Ubaldo, University of Sussex 35