Penny GoldbergPrinceton,
BREAD, NBER
Trade and Growth:What can we learn from micro
data?Sargan LectureApril, 21 2009
This talk draws on work with Amit Khandelwal (Columbia), Nina Pavcnik (Dartmouth) and Petia Topalova (IMF), whom I thank for their numerous insights and feedback.
Motivation
• Protectionists’ demands on the rise e.g., Recovery Act: “Buy American” clause forbids spending on
public works unless all of the iron, steel, and manufactured goods used are produced in the United States.
• Trade important for all countries, but particularly for developing countries that may have limited access to high quality intermediate products
• Indian policymakers:
“Trade liberalization and tariff reforms have provided increased access
to Indian companies to the best inputs available globally at almost world prices.”
–Rakesh Mohan, Deputy Governor of Reserve Bank of India (2008)
Trade and Growth: Mechanisms
• Efficiency Gains arising from Specialization and/or Selection• Benefits of Scale and Competition• Access to New Technology, Intermediate and Capital Goods.
On the other hand:
• Potentially Adverse Effects (e.g., specialization in less dynamic sectors that reduces long-run growth rate, infant-industry arguments, imperfect competition)
Ultimately: Empirical Question
Evidence
Evidence (largely based on cross-country studies) highly contentious
Reasons:
1. Measurement Problems (particularly regarding “openness”)2. Reverse Causality and Omitted Variable Bias3. Trade Policies need to be combined with other measures4. Most Studies focus on pre-1990 data and ignore recent
liberalization waves in developing countries
Critical Reviews of this Literature by: Rodriguez and Rodrik (2001), Hallak and Levinsohn (NBER Wo. Paper, 2004), Winters, McCulloch and McCay (JEL, 2004), Estevadeordal and Taylor (NBER Wo. Paper, 2008), and many others.
Focus
Focus (continued)
• Approach philosophically similar to the one used in sectoral studies of effects of trade liberalization on TFP
• Part of growing literature that exploits micro data in order to understand the determinants of growth. Examples include:
○ Hsieh and Klenow (2009): Allocative efficiency and growth○ Bloom and Van Reenen (2007): Importance of managerial
practices○ Bloom and Van Reenen (2008): Impact of trade (Chinese
imports) on innovation and information technology
• Here: Focus on the role of intermediate inputs rather than final goods trade. The former is arguably more important for the developing world.
Road Map
• Theory• Difficulties in bringing the theory to the data• What is different these days?
• More disaggregate data at the plant or firm level• Trade liberalization episodes in developing countries
• Evidence Based on India
Theory
• Endogenous Growth Models:○ Many different versions, but common theme: new varieties
play an important role in growth (Rivera-Batiz and Romer, Romer, Grossman and Helpman)
• Two Effects:
○ Level (static) effect: New Intermediate and Capital Goods lead to increases in measured TFP
○ Growth (dynamic) effect: Arises from a linkage between new variety creation and research activity. As productivity increases, the rate of variety creation will rise long-run growth rate increases
Level Effect
• Simplest possible model: Consider production function:
Under standard assumptions,
Output Growth:
and
Level Effect (continued)
• Love of Variety: Increase in N (through trade) leads to increase in TFP• Many reasons simple version empirically undesirable• More general models in empirical work still “variety” effect
remains• Effects operates WITHIN firms, not through selection.
• Most compelling empirical evidence so far:○ Amiti and Konings (AER): Reduced Form
○ Show reductions of INPUT tariffs in Indonesia TFP increase
○ Halpern, Koren and Szeidl (2008): Data on imported inputso Show imports of intermediates associated with higher TFP (no
liberalization)
• Potential Problem: Measurement of TFP Is it “efficiency” or “markup”? (De Loecker, 2008)
Growth Effect
Dynamic Gains from Trade
• Trade has two effects in the short- run:• Lowers prices of imported products price effect• Brings new products to the domestic country variety effect
• These two effects can by captured through the changes in the exact import price index. The exact price index has two components:
○ Conventional price index (captures price changes of existing products)
○ Variety component: captures the price change for NEW products from their virtual price (0 imports) to the price we observe upon entry
• Import of new intermediate products corresponds to a decrease in the exact price index of intermediates
Growth Effect (continued)
Implications for Growth:
• Consider a domestic firm facing a fixed cost of introducing a new product. The decrease in the price index for intermediates implies that the firm is more likely to introduce new domestic products.
• Feedback Effect: Some of these new products will be used as inputs in the production of other domestic goods. If new products = final goods, pos. externalities if fixed costs of production decreasing in N of products
• “Lab-Equipment” Model: Innovation requires capital or intermediate inputs. Since new imported intermediates lower the exact price index for intermediates, they lower the cost of doing research. This creates a DIRECT link between imported intermediates and R&D.
More formally, consider the following production technology:
Production Technology
• Production is CRTS and Cobb-Douglas
• Each input sector is comprised of domestic and foreign inputs that are combined using a CES aggregator
• The foreign input is also given by the CES aggregator:
Exact Input Price Index
• The exact price index associated with foreign input sector is:
where are the Sato-Vartia ideal weights and and ‘ denotes the previous time period
Minimum Cost Function
Mechanism (continued)
• In addition, direct effect on research activity of firms, if research requires capital and intermediate products.
Summary of Mechanism
• Trade Liberalization leads to lower prices for imported intermediate and capital goods as well as the import of new varieties of these goods
○ Intensive margin lower prices○ Extensive margin new goods (leads to higher measured TFP)
• As a result, the minimum unit cost associated with the production of a domestic good falls
• Domestic producers may now find it profitable to start producing products that were unprofitable prior to the liberalization.
• The introduction of new domestic varieties further contributes to growth
○ Domestic Intermediate Products: Same argument as with imported feedback effect
○ Domestic Consumer Products: Lead to further growth if:○ Externalities through fixed costs declining in the number of
goods○ Concave Transformation Curve
Difficulties in Bringing the Theory to the Data
• Lack of Information on Domestic Product Creation and R&D• Potential reverse causality between imports of new
intermediates and outcomes of interest
What is different now?
1. Better data (at the plant or firm level)2. Trade Liberalization Episodes in Developing Countries Provide
plausibly exogenous variation for identifying the effects of interest.
DataIndia: Prowess Database
• Center for Monitoring of Indian Economy (CMIE)○ Covers 60-70% of India’s economic activity in industrial sector.
• Firm-level panel data 1989-2003○ Spans the period of trade reforms
• Medium and large Indian firms○ Not well suited to study firm entry and exit○ Well suited to study product mix changes among mfg firms
• Mfg firms required by law to disclose product-level information in their annual reports (1956 Companies Act)
○ Product module contains annual product-level information○ Information on unit values
• Information on R&D
Data: Definition of a Product
• Based on CMIE’s internal product classification and products reported in companies’ annual reports
• 1,886 products○ Comparable to BRS (2006a) product definition (~1500 5-digit
SIC)○ Linked to 108 4-digit NIC industries
• Coverage across firms○ Product info available for vast majority (85%) of mfg firms○ 85% of manufacturing output
Examples of Products
India’s Trade Liberalization
• Twenty years ago, the infamous “license raj” required firms to obtain official permission to expand into new product lines and upgrade capacity.
• India’s pre-reform trade regime was very restrictive
• During the 1980s, India initiated a process of liberalization○ Import and industrial licensing eased, but tariffs remained very
high
• In August 1991, India faces a BOP crisis and calls the IMF○ Tariff changes 1992 - 97 largely set as part of IMF conditions in
1991
• Avg nominal tariffs fall from 90% in 1987 to 30% in 1997○ Unanticipated & unlikely to be foreseen in pre-reform industry
decisions Topalova (2007), Gang and Pandey (1996)
Source: Topalova (2007)
India’s Trade Liberalization
India’s Liberalization
Trade Data
• India’s import data (1987-2000) at the HS8-country level of disaggregation
• Tariff data is reported at the HS6-level
• Generate input tariffs using India’s input-output matrix○ Average input tariffs fall from 36% in 1987 to 12% in 1997
• Nourez (2001) classifies India’s products as intermediates (intermediates, basic & capital goods) or final goods
○ Note: will use this classification for descriptive tables & rely on IO matrix for empirical analysis
Examples of Consumer & Intermediate Products
(Import Data, Nouroz, 2001)Examples of Intermediate Products
HS 8504 Electrical transformers, static converters ( eg rectifiers ) & inductors
HS 8505 Electro-magnets; permanent magnets, etc; electric-magnetic or permanent magnet chucks & similar holding devices, etc
HS 8508 Electro-mechanical tools for working in the hand, with self-contained electric motorHS 8608 Railway or tramway track fixtures & fittings; mechanical signalling, safety or traffic control equipmentHS 8456 Machine-tools for working any material by removal of material, by laser or other lightHS 7002 Glass in balls, rods or tubes, unworkedHS 7017 Laboratory, hygienic or pharmaceutical glassware, whether or not graduated or calibratedHS 7019 Glass fibres ( incl glass wool ) and articles thereof ( eg yarn, woven fabrics )HS 7104 Synthetic or reconstructed precious or semi-precious stones, not strung or temporarily strung, not mounted or setHS 2901 Acyclic hydrocarbonsHS 2903 Halogenated derivatives of hydrocarbonsHS 28xx Inorganic and Organic chemicals and their compounds of precious metals.
Examples of Consumer ProductsHS 7113 Articles of jewellery & parts thereof, of precious metal or of metal clad with precious metalHS 8415 Air conditioning machines, comprising a motor-driven fan & elements for changing the temperature &HS 8521 Video recording or reproducing apparatus, whether or not incorporating a video tunerHS 8524 Records, tapes & other recorded media for sound or other similarly recorded phenommena, exclHS 62xx ClothingHS 64xx Footwear
Evidence
Arguments has three main parts:
A. Descriptive Part1) Indian firms add products after its large-scale trade reform GKPT
(2008). Net product additions account for 25 percent of growth in
India’s manufacturing growth in 1990s 2) Tariff liberalization is “real” surge in imports, particularly of
intermediate products. 2/3 of the growth in intermediate imports accounted for by
new varieties
B. Reduced Form Evidence Input tariff reductions lead to an expansion of domestic products
and R&D
C. Semi-Structural Part: Impose more structure in order to identify channels: Lower prices
versus increased variety.
A. Domestic Product Growth
• Extensive Margin important for growth new products account for one quarter of growth in manufacturing
• Net = Gross No product dropping
• Stark contrast to the U.S. and also other developing countries no evidence of “creative destruction”
• Explanations:○ Remnants of license-raj: High sunk costs associated with
earlier regulation○ Large developing country with huge income disparities
always demand for “older” products (consistent with the fact that there are few product categories that are “shrinking”)
• No concern that higher imports are displacing domestic products
A. Tariff Declines and Import Growth
• India’s liberalization had “real” impacts on India’s imports
• Within HS6 products, lower tariffs implied ○ Higher import volumes○ Lower unit values○ More import varieties (particularly for intermediate goods)
• The importance of variety growth is seen in the following decomposition table
• Messages of table:○ India experienced a surge in overall imports○ Intermediate imports increased by more than 200% during
the 1990s○ New products account for 2/3 of intermediate import growth
• Intermediate extensive margin growth driven by imports from OECD + E. Asian countries GKPT (2009)
○ Within HS6-Year pairs, higher unit values for new varieties
Decomposition of India’s Imports
Product TypeImport Growth Net
Product Entry
Product Exit Net Growing Shinking
All Products 130 84 84 0 45 -39 84
Intermediate Products 227 153 153 0 74 -42 116
Final Products 90 33 33 0 57 -29 86
Table 1: Decomposition of Import Growth, 1987-2000Extensive Margin Intensive Margin
Shrinking Growing
Variety Intensive Margin
NetGross, OECD
Gross, ROW Net
Gross, OECD
Gross, ROW Net
All Products 130 84 59 25 45 22 9 13 23Capital, Basic, Intermtds 227 153 115 38 74 42 15 26 32Final Products 90 33 21 11 57 25 9 16 32Basic Products 260 154 124 30 106 62 31 31 45Capital Products 125 37 27 10 88 33 23 10 55Intermediate Products 297 278 200 78 19 28 -12 39 -9
Product (HS6) Extensive Margin
Product (HS6) Intensive Margin
Variety (HS6-Country) Extensive Margin
NetImport Growth
Decomposition of India’s Imports, by Source
• Extensive margin growth driven new products/varieties from OECD (plus HK, Singapore, Taiwan)
• HS6 intensive margin driven by new varieties (HS6-country)
• We also observe higher unit values for added varieties within HS6-year
Notes: OECD includes Hong Kong, Singapore and Taiwan
New HS6 Codes
HS Code Examples of New HS Goods720890 Hot rolled iron or non-alloy steel, width >600mm, nes847192 Computer input or output units847191 Digital computer cpu with some of storage/input/output851740 Apparatus, for carrier-current line systems, nes847120 Digital computers with cpu and input-output units854800 Electrical parts of machinery and apparatus, nes854280 Electronic integrated circuits/microassemblies, nes854220 Hybrid integrated circuits382390 Chemical industry products, preparations, mixtures ne854030 Cathode-ray tubes, except for television720990 Cold rolled iron or non-alloy steel, flat, width >600mm, nes
310540 Monoammonium phosphate & mix with diammonium, <=10 kg
B. Reduced Form Relationship
• Why might declines in trade costs lead to expansion of firm product scope?
• Firms face a fixed cost of manufacturing a product. Lower variable costs due to the liberalization enable firms to
cover the fixed costs of entering new product lines
• Reduced-form specification relates firm scope to the input tariff faced by firms in sector s
○ Input tariff of industry line m○ Year fixed effects (t), firm fixed effects (f)○ Standard errors clustered at the industry level○ Restrict sample to 1989-1997
Expansion of Product Scope and Input Tariffs
• Avg input tariffs fell 24 pp 7.7 % increase in firms’ product scope ○ Accounts for about 31% of overall firms’ product scope
expansion• From GKPT (2008), new products account for 25% of India’s mfg
growth ○ This particular channel accounts for ~8% of overall mfg growth
Product Scope and Input Tariffs (1) (2) (3) (4) Input Tariff -0.323** -0.327** -0.309** -0.278** 0.139 0.140 0.138 0.117
Output Tariff
-0.037 -0.033 -0.035
0.035 0.039 0.032 Delicense -0.032 -0.027 0.023 0.021 FDI 0.035 0.024
Year Effects yes yes yes yes Firm FEs yes yes yes yes Adjusted R2 0.875 0.871 0.875 0.877Observations 14,872 13,435 14,854 11,135
Input Tariffs and Other Firm Outcomes
Log Firm Sales Log TFP R&D Indicator R&D Indicator
Lagged Input Tariff -0.987 ** -0.454 * -0.293 0.1940.414 0.233 0.268 0.271
Lagged Input Tariff -0.629 ***
x Firm Size Indicator 0.111
R-squared 0.93 0.81 0.52 0.59
Observations 14,204 13,714 14,233 14233
• Lower input tariffs associated with higher• Output• Productivity (note: TFP taken from Topalova (2007))
• Probability of conducting R&D among the (initially) large firms
Source: Regressions include firm and year fixed effects
C. Mechanism
• So far: Strong and robust relationship between input tariffs and product scope expansion, R&D, sales and TFP.
• Impose additional structure to assess the relative contribution of lower existing input prices and firms’ access to new imported input varieties
• Rely on production technology described above to decompose exact import price index into a price and variety channel
Minimum Cost Function
• Explicit assumptions on the structure
• The Cobb-Douglas weights are the coefficients from India’s input-output matrix
• Impose proportionality in the input-output coefficients between domestic inputs and foreign inputs
Next Steps
1. Compute the exact price index, conventional price index and variety component for each sector.
2. Run those through Input-Output tables to create input price indices.
3. Relate changes in firm product scope to the changes in the exact price index as well as its components
Variety Index
• Mean variety index for intermediates is 0.881 between 1989-1997 (A lower variety index implies larger gains from variety)
• Right column reports the aggregate variety component of the import price index from 1989-1997
• Aggregate exact import price index for intermediates is about 4.7% lower per year due to new varieties
• These variety gains are substantially larger than those reported for U.S. and Costa Rica Broda & Weinstein (2004), Arkolakis et al (2008)
Mean Variety Index Overall Variety Index
All Products .899 .688
Intermediates .881 .624
Final Goods .904 .850
Input Indexes
• Using the IO tables, construct input indexes for each sector
• Relate firm product scope to changes in each index
Input Tariffs and Imported Input Price Index
Log Exact Price Index
Log Conventional Price Index Log Variety Index
Input Tariff 0.28 *** 0.132 *** 0.148 ***
0.051 0.05 0.014
R-squared 0.18 0.05 0.47Observations 136 136 136
Long difference regression: 1989 and 1997.
Decomposing Channels(OLS Regressions)
Long difference regression: 1989 and 1997.
Log (Number of Products)
Log Exact Price Input Index -0.4910.343
Log Conventional Input Price Index -0.0090.485
Log Variety Input Index -2.615 **
1.186
R-squared 0.00 0.01Observations 696 696
Identification Assumptions
• Explicit assumptions on the structure
• Error term certainly correlated with right-hand side variables.
• Identification assumption:○ Input tariffs affect domestic prices and final sector TFP only
through the import channel
Instruments
• Tariffs: plausibly exogenous in this setting and highly correlated with variables on the right hand side (see above, tariffs affect unit values and range of imported intermediates).
• Country–specific fixed costs of entry into the Indian market (variables suggested by gravity-type models, such as language) interacted with measures of sector-specific comparative advantage Idea:
• Conditional on changes in tariffs, sectors in which potential exporters face a lower fixed cost of exporting to India will experience differential changes in the new varieties imported
• Preliminary results stronger than OLS results Domestic product scope expansion due to firms’ access to new intermediate inputs. Same is true for R&D.
Conclusions
• New products account for 25% of India’s mfg growth in 1990s
• Imports of intermediates surge with 2/3 of growth due to new products
• Connect these two facts o Firms’ access to new inputs from abroad leads to product
scope expansiono New inputs associated with R&D
• Empirical support for the microeconomic mechanism of domestic variety expansion in endogenous growth models
Final Remarks
• Mechanism not unique to India. ○ Anecdotal Evidence from Coca-Cola Factory in Shanghai
• Reminder: Trade is about imports, not just exports.○ Paul Krugman: “Moreover, the purpose of international trade
-- the reason it is useful -- is to import, not to export. That is, what a country really gains from trade is the ability to import things it wants. Exports are not an objective in and of themselves; the need to export is a burden that a country must bear because its import suppliers are crass enough to demand payment.”
THANK YOU!