Distribution Capital and the Short and Long Run Import Demand Elasticity M.J. Crucini and J.S. Davis

Transcription

1 Distribution Capital and the Short and Long Run Import Demand Elasticity M.J. Crucini and J.S. Davis Discussant: Andrea Rao Board of Governors of the Federal Reserve System

2 CD (2012): Motivation The trade elasticity (") is a key (and controversial) parameter for open economy models. - Trade models: We observe large increases in exports and imports after trade liberalizations (4 < " < 10). - IRBC models: We observe large uctuations in relative prices over the the business cycle (0:5 < " < 1:5). Can we reconcile these two facts?

3 CD (2012): Yes, We Can! (1) Model Standard IRBC model a la BKK augmented to include a distribution sector. - Idea: you need Amazon.com and Amazon.it to deliver Californian and Italian olive oil to Washington DC. Role of distribution sector is not new, but authors use micro data to pin down key shares and elasticities. - Data require Leontief aggregation in distribution (;! 0).

4 CD (2012): Yes, We Can! (2) Findings 1. Trade elasticity is low in the short run but high in the long run " t = (1 s) + f t where f t < 0 in the short run but eventually converges to zero. 2. Model performs quantitatively well (business cycle stats, S-curve).

5 Quarters Figure 1: Observed Elasticity of Substitution following a TFP shock. The solid line is where the structural elasticity is equal to 4 and there is no distribution. The dashed line is where the structural elasticity is equal to 0.9 and there is no distribution. The line with stars is where the structural elasticity is equal to 4 and there is distribution. Table 1: Parameter Values Symbol Value Description 0:66 weight on leisure in the household s utility function 2 coe cient of relative risk aversion 0:36 capital share in the production of traded goods 0:99 discount factor! 0:85 exogenous preference for home goods 0:025 capital depreciation rate 0:0001 elasticity of substitution between wholesale goods and distribution services d 0:90 weight on capital in the production technology for distribution services 0:1 elasticity of substitution between capital and labor in distribution 0:375 capital adjustment cost parameter for capital used in distribution 27

6 What Makes a Good Discussion You provide authors with constructive feedback. Avoid 1. promoting your own work; 2. questioning the assumptions. I will violate both recommendations.

7 Outline of the Discussion 1. Detour: Quick overview of data and theory. 2. Quantitative performance: success? 3. Micro data and calibration.

8 Quick Overview of Data and Theory (1) In the data, Net Exports (NX) are countercyclical - countries borrow in good time (against C-smoothing). - Investment drives NX dynamics ("Make hay while the sun shines"). Intertemporal approach to current account: NX used to smooth cycles. BKK success: model is able to account for this key feature of the data, preserving business cycle regularities.

9 Table 1. Correlation between Net Exports and Output Developed Economies Emerging Economies Australia 0.36 Argentina 0.89 Belgium 0.18 Brazil 0.03 Canada 0.17 Ecuador 0.79 Finland 0.27 Israel 0.12 France 0.41 Korea 0.86 Germany 0.07 Malaysia 0.74 Greece 0.39 Mexico 0.87 Italy 0.27 Peru 0.24 Japan 0.40 Philippines 0.40 Netherlands 0.15 Slovak Republic Norway 0.01 South Africa Spain 0.38 Thailand 0.83 Sweden 0.04 Turkey 0.69 Switzerland 0.21 US 0.49 UK 0.52 EU Source: Raffo (2008). Average 0.41 Median 0.39

10 Quick Overview of Data and Theory (2) In the data, international prices and trade ows are quite volatile over the business cycle. BKK model delivers very little volatility on both dimensions. Key insight from the theory: - Armington structure creates a tradeo between volatility of trade prices and quantities; - This tradeo depends critically on the trade elasticity ("):

11 Table 1. Volatility: Net Exports and Terms of Trade Standard Deviation x Net Exports Terms of Trade Australia Belgium Canada Denmark Finland France Germany Italy Japan the Netherlands New Zealand Portugal Spain Sweden United Kingdom United States Median BKK [1994] - Benchmark High Elasticity x Standard deviations relative to the standard deviation of GDP. Statistics refer to HP-ltered quarterly data for the period 1980Q1-2007Q4. 43

12 2.5 2 Standard Deviation Elasticity of Substitution Terms of Trade Import Ratio

13 Quantitative Performance of the Model (1) How does the CD (2012) model performs in terms of the international price and quantities? How does the CD (2012) model performs in terms of the standard business cycle properties?

14 Quantitative Performance of the Model (1) How does the CD (2012) model performs in terms of the international price and quantities? - (Guess): NX is not countercyclical, X and M do not move; - STD(TOT) is, not suprisingly, as high as in the data. How does the CD (2012) model performs in terms of the standard business cycle properties? - STD of I and L are low.

15 Table 2. Quantitative Performance of the CD (2012) Model: Success? U.S. Data BKK CD (2012) STD(X)/STD(GDP) C I N TOT NX Corr (X, GDP) C I N TOT NX Corr (TOT, NX) NB. BKK as in Raffo (2008).

16 Table 2. Quantitative Performance of the CD (2012) Model: Success? U.S. Data BKK CD (2012) STD(X)/STD(GDP) C I N TOT NX Corr (X, GDP) C I N TOT NX Corr (TOT, NX) NB. BKK as in Raffo (2008).

17 Quantitative Performance of the Model (2) CD (2012): NX is weakly countercyclical and probably for the wrong reasons (tot reects relative scarcity) cnx = im h \nxqty dtot i Rao [2008]: In the data, nxqty is countercyclical and models often cannot reproduce this (I adjustment costs, trade shares,...). Advice: use decomposition nxqty vs tot (data, model statistics, IRF).

18 Table 2. Quantitative Performance of the CD (2012) Model: Success? U.S. Data BKK CD (2012) STD(X)/STD(GDP) C I N TOT NX Corr (X, GDP) C I N TOT NX Corr (TOT, NX) NB. BKK as in Raffo (2008).

19 Quantitative Performance of the Model (3) CD (2012): Volatility of trade prices is high, but "! 0 in the short run ln(tot) = + 1 " ln B A Hence trade quantities do not move. In the data, both trade prices and quantities are quite volatile. Can the model quantitatively reconcile trade and IRBC models?

20 Table 2. Quantitative Performance of the CD (2012) Model: Success? U.S. Data BKK CD (2012) STD(X)/STD(GDP) C I N TOT NX Corr (X, GDP) C I N TOT NX Corr (TOT, NX) NB. BKK as in Raffo (2008).

21 Micro Data and Calibration (1): Comovement Data: pervasive comovement across sectors (output, labor, asset prices...). In CD (2012), you get - small adjustment in quantities and large changes in prices (Leontief); - small response in aggregate labor (reallocation across sectors); - large dierentials in r k across sectors and negative comovement in investment. Advice: provide evidence that quantities, labor patterns, and/or asset prices in the data conform to the model (discipline).

22 Micro Data and Calibration (2): Durable Goods In the data, about half of trade involves durables (E&S...), not nal goods. - Engel and Wang (2011): machinery and transport equipment account for 40% of imports across OECD countries. - This evidence is often used to motivate the high variability in trade ows over the business cycle. CD (2012) use micro data on retail goods to parametrize the model. Appropriate (Leontief)?

23 Micro Data and Calibration (3): How Fixed is Capital? CD (2012): in the short run, distribution capacity is xed across sectors. Amazon.com and Amazon.it to deliver California and Italian olive oil to Washington DC - Can Amazon reallocate capacity?

24 Final Remarks CD (2012) is an ambitious and promising paper: use microevidence on retail and distribution margins to inform about short and long run trade elasticity. Advice: pay closer attention to the macroeconomic implications of the model. I look forward to reading the next version!