International Trade and Development

1
International Trade and Development
2
International Trade and Development

• Lecture Outline
• (1) What do we include in a Growth model?
• (2) Evidence of the relationship between
  increased trade (globalisation) and growth
  relate to evidence of poverty/inequality and
  development.

3
International Trade and Development

• (1) What do we include in a Growth model?
• Based on the work of Levine and Renelt (1992,
  AER, 82(4)).
• Prior to this paper, empirical papers were not
  controlling for appropriate factors.
• Levine and Renelt found that many explanatory
  variables were not robust to changes to the
  inclusion/exclusion of other variables some
  even changed sign!!!

4
International Trade and Development

• (1) What do we include in a Growth model?
• Following a survey of empirical work LR decide
  upon 4 explanatory variables,
• where INVinvestment share of GDP
  is initial level of GDP per capita in 1960 (when
  data begins from) SEC is the initial
  secondary-school enrollment rate GPOP is the
  average annual growth rate of the population
  growth.

5
International Trade and Development

• (1) What do we include in a Growth model?
• (Q) Why these variables?
• Expect higher initial GDP level in 1960 to have
  a negative effect on real GDP/capita growth
  testing the hypothesis that a poor country,
  ceteris paribus, tends to grow faster than a rich
  country.

Growth of initially high GDP country
GDP
Growth of initially low GDP country
Year
6
International Trade and Development

• (1) What do we include in a Growth model?
• Expect higher initial level of human capital to
  positively effect real GDP/capita growth
  consistent with Solow growth model and more
  importantly the endogenous growth models.
• Increase in INV/GDP expected to positively
  effect real GDP/capita growth tests the Solow
  model and any other economic growth model.
  Investment is a key driver of growth.
• Those countries with high rates of population
  growth have lower growth rates finding is not
  robust to all models.

7
International Trade and Development

• (2) The Empirical Debate Trade and Development
• We will discuss a number of papers on, the
  relationship between trade and GDP, on the effect
  of trade openness on GDP and finally the effect
  of trade on poverty and inequality.
• Look at Greenaway et al (2002), Irwin and Tervio
  (2000) and Dollar and Kraay (2004).
• First 2 papers focus on the impact of
  international trade on growth/income of countries
  using samples of countries over a number of
  years.
• Dollar and Kraay (2004) focuses on link between
  trade and poverty/income inequality.

8
International Trade and Development

• The key concern though when estimating the link
  between trade and growth is the issue of
  endogeneity which prevents a correct estimate of
  the impact trade has on growth I.E.
• THE POSITIVE CORRELATION BETWEEN TRADE AND
  INCOME COULD MEAN THAT COUNTRIES WITH HIGHER
  INCOMES ENGAGE IN MORE TRADE RATHER THAN
  COUNTRIES WITH MORE TRADE HAVING MORE INCOME
  (Taken from Irwin and Tervio, 2000).

9
International Trade and Development

• Greenaway et al (2001)
• Focus on the issue of trade liberalisation and
  its impact on GDP growth so not looking at
  impact of trade liberalisation on poverty or
  income distribution or development.
• Reason for the paper is the inconclusive
  evidence from previous work argue that reason
  for debate is (i) inappropriate methodology and
  (ii) how trade liberalisation is measured.
• Previous research can be criticised for not
  using panel techniques that control for (i)
  country-specific effects, (ii) time-effects and
  (iii) an unobserved error effect that varies
  across both countries and years (time) which is
  assumed to be uncorrelated.

10
International Trade and Development

• Greenaway et al (2001) cont
• The Core growth model is based on the empirical
  work of Levine and Renalt (1992) and theoretical
  developments by Romer (1990).
• Robust variables included in most core growth
  models are
• (i) investment, (ii) population growth, (iii)
  initial human capital, (iv) initial GDP per
  capita.
• Greenaway et al also add to this list of
  explanatory variables, (v) terms of trade
  variable and (vi) trade liberalisation proxies.

11
International Trade and Development

• Greenaway et al (2001) cont
• The base specification is,

12
International Trade and Development

• Greenaway et al (2001) cont
• Problem with this specification is that it is
  inappropriate if looking for dynamic effects.
  Include lagged changes in real GDP/capita growth
  as well.
• Expect an improvement in terms of trade to
  positively effect real GDP/capita growth.
• Importantly for the paper, is testing the
  hypothesis of whether trade liberalisation
  effects real GDP/capita growth in any way. If
  effect ve then supports the pro-trade
  liberalisation group.

13
International Trade and Development

• Greenaway et al (2001) cont
• Issue of how to proxy trade liberalisation, with
  the suspicion that different proxies can have
  different effects.
• (1) Use a before and after Structural Adjustment
  Loan (SAL) variable (simple binary variable) as
  this loan at least signal intent of trade
  liberalisation.
• (2) Timing of liberalisation is measured using
  actual information on levels and changes in
  tariffs, quotas, exchange rate misalignment and
  export impediments/promoters. Through this can
  identify the year when liberalisation has taken
  place. Is modelled by a binary variable as is
  the SAL based on Dean et al (1994).

14
International Trade and Development

• Greenaway et al (2001) cont
• (3) A measure of closed-ness and openness in
  trade in terms of average tariff level. If
  average tarifflt39 then open, if above 39 then
  closed. Very arbitrary but even Sachs and Warner
  (1995) who come up with the measurement recognise
  this Does just give another measure of
  openness/close-ness though.
• Results.

15
International Trade and Development

• Greenaway et al (2001) contUsing Sachs and
  Warner

16
International Trade and Development

• Greenaway et al (2001) contusing Dean et al

17
International Trade and Development

• Greenaway et al (2001) contUsing SAL

18
International Trade and Development

• Greenaway et al (2001) cont
• Finding is that when the simple dummy trade
  liberalisation terms are included that in
  post-liberalisation period growth in real
  GDP/capita is greater than in the
  pre-liberalisation period.
• Implication is that trade liberalisation (all 3
  definitions used) is a good thing for growth but
  the effect is not that great.
• However, the simple dummy variable approach
  represents an average effect over a number of
  years. By switching on the liberalisation proxy
  for the year liberalisation is introduced only
  and then turning it off again we get a direct
  growth impact estimate in the first year only,
  with lags picking up the impact of reforms in
  subsequent years.
• There are however indications of first and
  second order serial correlation!!!

19
International Trade and Development

• The second-order serial correlation indicates a
  mis-specified model.
• Standard way to get around this issue is to
  include instruments in the model which here takes
  the form of lagged changes in real GDP/capita
  from Columns 3 in the above tables we see 2nd
  order serial correlation is resolved not
  significant
• This is an example of the instrumental variables
  (IV) approach
• Column 3 represents the best model in the paper.
  The signs on the explanatory variables are all
  expected and most remain significant.
• There is an element of sensitivity regarding
  which trade liberalisation measure is used as to
  its impact on growth in real GDP/capita.
• However, the evidence does give a more
  consistent picture that trade liberalisation has
  a positive impact on growth which may take time
  to work through (the lagged trade liberalisation
  terms) into a growth impact.

20
International Trade and Development

• Irwin and Tervio (2002)
• Flag the endogeneity issue of trade and GDP and
  the use of an instrument that is expected to
  effect trade but not GDP.
• The instrument is relatively new to the
  literature and was first used by Frankel and
  Romer (1999).
• They construct an instrument that takes into
  account the geographical location of a country
  and in particular the countrys distance from
  trading partners.

21
International Trade and Development

• Irwin and Tervio (2002) cont
• They go through the process of constructing the
  instrument itself by OLS regression.
• Actual trade of country i with country j as
  a share of GDP of country i is regressed onto a
  number of factors that includes distance between
  country i and country j.
• This regression will give predicted values for
  each country i trade with country j/GDP
  ratio.
• Explanatory factors include populations of the
  countries, area of the two countries, whether
  landlocked or not.

22
International Trade and Development

• Irwin and Tervio (2002)cont
• For each bilateral trade between country i and
  country j can be calculated.
  These are then summed to yield a predicted value
  of country is trade share with all sample
  countries j this is the instrument.
• From this prediction we construct the instrument
  necessary to solve the endogeneity problem in the
  growth equation.
• If the growth model uses (actual trade/GDP) then
  the coefficient in the growth equation on this
  term will be biased upwards.
• Reasons are things like richer countries being
  able to afford better infrastructure that means
  more trade is physically possible (e.g. ports,
  airports, road structure, railway structure).
• Technically it means that there is a positive
  correlation between the error term in the growth
  equation and the trade term.
• Need to use IV estimation to get over this issue
  this is where the constructed instrument comes
  into play!!

23
International Trade and Development

• Irwin and Tervio (2002)cont
• Instead of the growth equation being estimated
  through OLS, they use a two stage least squares
  (2SLS) approach.
• The T hat term represents the predicted trade of
  country i, that is a function of geographical
  distance between trading partners.
• In the first stage of this model, trade/GDP is
  regressed onto the instrument and a couple of
  other variables, i.e.
• The predicted values of trade from this first
  stage are calculated for each observation,
  represented by . This essentially
  represents a constructed trade share which is
  determined in large part by the bilateral trade
  estimate.

24
International Trade and Development

• Irwin and Tervio (2002)cont
• The final growth equation to be estimated is
  represented by,
• where the trade term is constructed from our
  regressions and does NOT represent actual
  trade/GDP weve determined what causes trade
  using a variable that cannot theoretically effect
  GDP.
• Compare the coefficient on the trade term with
  that for the coefficient in the OLS growth
  estimation represented by,

25
International Trade and Development

• Irwin and Tervio (2002) regression used to
  calculate the instrument

26
International Trade and Development

• Irwin and Tervio (2002) Growth models, OLS and
  2SLS using IV

27
International Trade and Development

• Irwin and Tervio (2002)cont
• Conclusions are that the effect of trade on
  income/capita of a country is larger using 2SLS
  than OLS (the coefficients on the trade variable
  are larger) although frequently less
  significant.
• Conclusion is that trade does positively effect
  growth even taking into account the endogeneity
  issue.
• Criticism of this type of model follows from
  Rodrik and Rodriguez (2000) with the model NOT
  robust to including distance from equator when
  Irwin and Tervio include this variable in the GDP
  equation the trade term becomes everywhere
  insignificant with the coefficient frequently
  negative!
• (Q) Is there an economic explanation for why
  latitude can be included in growth equations?

28
International Trade and Development

• Dollar and Kraay (2004) paper on Trade and
  Development, Poverty and Inequality
• Initially estimate a difference model based on
  the standard growth equation
• where LHS is log-level of per capita GDP in
  country c at time t. Could use log-level GDP,
  but then would clearly not be taking into account
  population size.
• is log-level of per capita GDP k years ago
  (the lag of growth in previous period).

29
International Trade and Development

• The matrix X is a set of control variables
  measured in averages. Volume of trade
  (exportsimports as a of GDP) is included in
  X.
• The 3 error terms are (i) an unobserved country
  effect that does not change over time, (ii) an
  unobserved error term that changes over time but
  which is common to all countries (e.g. oil
  prices), (iii) an unobserved error effect that
  varies across both countries and years (time)
  which is assumed to be uncorrelated.

30
International Trade and Development

• Most studies do not concentrate on levels of GDP
  but instead on changes in GDP over periods of
  time.
• This represents a regression of growth in GDP
  onto lagged growth in GDP and on changes in the
  set of explanatory variables in the X matrix and
  is represented formally by,

31
International Trade and Development

• The differenced growth equation has a number of
  desirable characteristics
• (1) whilst levels of trade (volume of trade)
  reflect more the geography of a country, changes
  in trade volume reflect something else since the
  geography of a country does not change.
• (2) institutional characteristics of a country
  also do not change considerably over time (e.g.
  racial composition, historical legacy of
  colonialism etc)
• (3) The difference growth model allows
  appropriate lags of the right hand side
  variables as instruments (Dollar and Kraay,
  2004, f38).

32
International Trade and Development

• The identifying assumption is that while trade
  volumes may be correlated with lagged shocks to
  GDP growth, trade volumes are not correlated with
  future shocks to GDP growth.
• In practice, this means that when we regress
  growth in the 1990s on(to) (i) growth in the
  1980s and (ii) the change in trade volumes
  (changes in trade as of GDP) between the 1980s
  and 1990s, we can use the level of trade volumes
  in the 1970s as an instrument for trade
  openness, (Dollar and Kraay, 2004, f39),
  brackets and italics added.

33
International Trade and Development Table from
Dollar and Kraay (2004) Standard Errors in
Brackets
  OLS IV IV IV 

Initial Income 0.419 0.783 0.765 0.96
  (0.071) (0.297) (0.367) (0.397)
Changes in Trade Volume 0.252 0.475 0.514 0.543
  (0.095) (0.175) (0.187) (0.210)
Contract Intensive Money     0.232  
      -0.41  
Government Consumption/GDP       -1.164
        (-1.009)
log(1Inflation Rate)       -0.142
        (-0.152)
Revolutions       -0.025
        (-0.084)
F-Statistic for First-Stage Regressions        
Lagged Growth Openness   12.46 17.49 8.09 16.17 8.56 10.62
Observations 187 187 153 173

34
International Trade and Development

• Interpretation of Table
• The point estimate on Trade volume in the OLS
  reads that a 100 increase in trade share of GDP
  would raise GDP in the country by 25 over a
  decade.
• When controlled for instruments on trade volume
  (See the significant F-Statistics for Lagged
  Growth and Openness), the size of the trade
  volume coefficient increases to near 0.5!!!
• Conclusion Greater involvement in trade is
  related to faster growth in developing nations.

35
International Trade and Development

• Cautionary Comment
• Researchers want to test whether lowering trade
  barriers to international trade significantly
  increases growth, once other factors have been
  controlled for implications if evidence
  supports this hypothesis is that governments
  should dismantle their barriers to trade
  (Rodrik and Rodriguez, 2000, pp. 3).
• Remember Rodrik and Rodriguez (2000) argue that
  applied pieces of work are using inappropriate
  indicators of trade policy to systematically
  bias quantitative results in favour of finding a
  statistically significant relationship between
  trade and growth.
• Dollar and Kraay (2004) use change in the volume
  of trade as a regressor in the change in growth
  model and identify the model by assuming that
  level of trade has no correlation with change in
  GDP in the future.
• However is this an appropriate identifier? Can
  this assumption be made?
• Remember, GDP and trade are clearly determined
  by one another and there is question of finding a
  variable(s) that determines change in trade but
  not a change in GDP.

36
International Trade and Development
37
References

• Levine, R., and Renelt, D., (1992), A
  Sensitivity Analysis of Cross-Country Growth
  Regressions The American Economic Review, Vol.
  82, No. 4 (Sep., 1992), pp. 942-963
• Dollar, D., and Kraay, A., (2004), Trade,
  Growth and Poverty, Economic Journal, Vol 114,
  pp. f22-f49.
• Frankel, J., and Romer, D., Does trade cause
  growth?, American Economic Review, Vol 89, pp.
  379-399.
• Greenaway, D., Morgan, W., and Wright, P.,
  (2002), Trade liberalisation and growth in
  developing countries, Journal of Development
  Economics, Vol 67, pp. 229-244.
• Irwin, D.A., and Tervio, M., (2002), Does trade
  raise income? Evidence from the twentieth
  century, Journal of International Economics, Vol
  58, pp. 1-18.
• Rodriguez, F., and Rodrik, D., (2000), Trade
  policy and economic growth a skeptics guide to
  the cross-national evidence, Working Paper
  (available upon request).
• Winters, A., (2004), Trade liberalization and
  economic performance An overview, Economic
  Journal, 114 F4-F21.
• See Dani Rodriks website at http//ksghome.harva
  rd.edu/drodrik/