Clemson University TigerPrints All Dissertations Dissertations 7-2013 BANKING CRISES AND THE VOLUME OF TDE Yifei Mu Clemson University, [email protected]Follow this and additional works at: hps://tigerprints.clemson.edu/all_dissertations Part of the Economics Commons is Dissertation is brought to you for free and open access by the Dissertations at TigerPrints. It has been accepted for inclusion in All Dissertations by an authorized administrator of TigerPrints. For more information, please contact [email protected]. Recommended Citation Mu, Yifei, "BANKING CRISES AND THE VOLUME OF TDE" (2013). All Dissertations. 1190. hps://tigerprints.clemson.edu/all_dissertations/1190
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Clemson UniversityTigerPrints
All Dissertations Dissertations
7-2013
BANKING CRISES AND THE VOLUME OFTRADEYifei MuClemson University, [email protected]
Follow this and additional works at: https://tigerprints.clemson.edu/all_dissertations
Part of the Economics Commons
This Dissertation is brought to you for free and open access by the Dissertations at TigerPrints. It has been accepted for inclusion in All Dissertations byan authorized administrator of TigerPrints. For more information, please contact [email protected].
Recommended CitationMu, Yifei, "BANKING CRISES AND THE VOLUME OF TRADE" (2013). All Dissertations. 1190.https://tigerprints.clemson.edu/all_dissertations/1190
Where the utility of consumers in country j is, is the good consumed by people in
country j imported from country i, is the elasticity of substitution and 5 .
4 See Anderson and Van Wincoop (2003) for details.
5 The preference exhibits “love of variety”.
6
Maximizing utility subject to a budget constraint can solved out the demand for the
good consumed in country j import from country i, :
. (1.2)
Here is the price of the good’s price sold within the importer i. is the trade cost6 for
good shipped from country i to country j. is GDP of country j, and is the CES price
index that:
. (1.3)
Assume firms maximize profit and all markets clear, we can write an expression for
bilateral trade flow as:
. (1.4)
Where is the world gross GDP and
. (1.5)
. (1.6)
denotes , that is the share of country i’s GDP relative to the world. and are
usually known as multilateral resistance. is the outward multilateral resistance which
measures how difficult for country i to export goods relative to the rest of the world.
6 We can assume the trade costs as iceberg trade costs, the cost for goods were lost in transit.
7
is the inward multilateral resistance that measures how difficult for country j to import
goods relative to the rest of the world. Anderson and Van Wincoop (2003) noticed that
when influence trade is estimated, it is critical to include both inward and outward
multilateral resistances into the regression.
Taking the natural log of both side of equation 1.4, we can get:
. (1.7)
This measures relation between trade flow on the left side and trade cost, multilateral
resistance, and GDP on the right side with in some time period. When the trade cross
some time periods, equation 1.7 can be presented as:
. (1.8)
1.3.2 Bilateral effects estimation
In this paper, trade cost contains two components. One is traditional geography
variables, like distance, contiguity and common official language, which were wildly
used in lots of research. These variables are bilateral relations and are time invariant.
Other factors may be time varying including whether a country has a banking crisis.
More specifically, we assume the trade cost has the structure as follow:
. (1.9)
Where is the distance between country i and j, is the vector include other
geography information, like contiguity and language. is the vector contains the
8
information about financial shocks and can be presents in several different forms.
Replace equation 1.9 into equation 1.8, we get:
. (1.10)
is the error term with normal distribution.
The regression will include country-year fixed effects7. These fixed effects will
absorb the inward/outward multilateral effect effects, importer’s GDP, exporter’s GDP
and world GDP for the same year. We will also allow the financial shocks to have a
bilateral effect. The regression we are going to estimate is
. (1.11)
Where is the exporter-year is fixed effect and is the importer-year fixed effect.
We use the data from 173 countries across 32 years, so the number of dummy
variables for fixed effect will surpass 11000. If importer-exporter fixed effects are also
included, the number of dummy variables will surpass 25000, yield a large computational
burden. Also, the inverse a large matrix is usually imprecise. Previous method of
7Both importer-year fixed effect and exporter-year fixed effect.
9
reducing the computational burden has been to include only one out of every three years
of data with which to estimate regressions. Guimaraes and Portugal (2009) suggest a
method to estimate the models with high-dimensional fixed effects. This paper uses this
algorithm to make it possible to estimate the influence of banking crises while including
all data and all relevant fixed effects. The results are showed in Table 1.5.
1.3.3 Unilateral effects estimation
When the financial shock is treated as unilateral effect, trade costs are defined as:
. (1.12)
Baier and Bergstrand (2009) introduced the method to linear approximate the
multilateral resistances. For the bilateral trade costs, those resistances term in equation
1.5 and 1.6 can be presented as:
. (1.13)
. (1.14)
Plug equation 1.13 and 1.14 back to equation 1.8 and take the linear expansion to
, combine with equation 1.12, The regression changed into:
. (1.15)
10
Where:
. (1.16)
. (1.17)
. (1.18)
. (1.19)
. (1.20)
. (1.21)
is the measurement for contiguity, is the measurement for common
language. Since , and
are kind of the gross average of the
distance, contiguity and language of the world, these variables are constant for the same
year. Use year fixed effect will absorb all these variables and , which is the GDP
for the whole world at that year. The regression will be estimated is:
. (1.22)
Here is year fixed effect. Table 1.3 and 1.4 present the results for the unilateral
effects8.
8 For unilateral effect, The term
,
11
1.3.4 Unilateral effects estimation with two-stage model
Another way to think about the unilateral financial shock effects is that all these
effects have the same impact across all exporters (or importers). In this case, we employ a
two-step process to back out the effects of the banking crisis on exporters and importers.
In first-stage, we run the regress the log of bilateral trade on all trade costs except .
We then use the coefficients for and that captures all the country specific
information, include log of GDP and the multilateral resistance. For the exporter-year
fixed effect coefficient, it contains unilateral financial shock effects from exporter,
importers’ GDP share weighted bilateral effects9, some of the average trade cost cross the
world10, exporter’s GDP and some of the world GDP. So we can run the second stage
regression as follows:
.
(1.23)
From equation 1.8, the theory tells us the coefficient of log exporter’s GDP should be
equal to -1 when moved to left side. will capture all the information from bilateral
effects, world average trade cost effects and world GDP for the same year.
it actually becomes
.
Since the share of GDP sum up equal to one cross the world. and
is constant
for all the countries in the same year. Because of this unilateral effect cannot do the same expansion as
bilateral effects. We just include the unilateral effects by themselves. 9 Which is
.
10 Which is
.
12
Similar as previous, for importer-year fixed effect, we run the second stage
regression as:
.
(1.24)
Results are presented in Table 1.6 to 1.8, these also can be used a robustness
check compare to the results from equation 1.22.
1.4 Data Source
1.4.1 Value of the Trade
The value of bilateral trade for 173 countries for the years 1976-2008 is taken
from the UN Comtrade database11
. It is reported as 5 digit SITC level and aggregated as
each country’s import and export value. The value is measured in thousands of US dollar
in the current year. The inflation of the currency will be captured by importer-year fixed
effect and exporter-year fixed effect.
1.4.2 Geography Data
Geography data is used to measure the traditional trade cost. The paper uses
bilateral value of the distance, contiguity and common official language as measures of
the traditional trade cost. The data are from CEPII database12
. Both contiguity and
common language are dummy variables. Contiguity is equal to unity if two trade partners
11
The data can be obtained from http://comtrade.un.org/ 12
The data can be obtained from http://www.cepii.fr/CEPII/en/bdd_modele/bdd.asp
13
share the common border, and zero otherwise. Common official language is equal to
unity if two trade partners use the same official language and zero otherwise. CEPII
provides both simple great circle distance and population weighted distance between
countries. This paper uses the population weighted distance13
.
1.4.3 Banking Crises Data
Our data on banking crises is from the Leaven and Valencia banking crises
database provides annual banking crisis data for the year 1976-200814
. Leaven and
Valencia (2012) define a banking crisis as systemic if two conditions are met:
1) Significant signs of financial distress in the banking system (as indicated by
significant bank runs, losses in the banking system, and/or bank liquidations)
2) Significant banking policy intervention measures in response to significant
losses in the banking system.
Here, significant bank runs indicate a 5 percent or greater drop in deposits within
one month during the time period.
For policy interventions in the banking sector to be significant, at least three out of
the following six measures must have been used:
1) extensive liquidity support (5 percent of deposits and liabilities to nonresidents)
2) bank restructuring gross costs (at least 3 percent of GDP)
13
Use other measurement of distance will yield similar results. 14
The data can be obtained from http://www.imf.org/external/publications/index.htm
14
3) significant bank nationalizations
4) significant guarantees put in place
5) significant asset purchases (at least 5 percent of GDP)
6) deposit freezes and/or bank holidays.
In this paper, for bilateral effects, crises data are used if either one country or both
trade partners are experiencing a crisis. The crises lag variables capture the impact on
trade for the years after banking crises. In order to clarify whether bilateral trade is
impacted pre-crisis, forward crises variables will also be used.
In total, the panel dataset contains trade vales, banking, and geographic data for
173 countries for the years 1976-2008.
1.5 Results
As we mentioned in introduction, there may be two channels through which a
banking crisis might influence a trade flows. If the banking crisis’s impacts occur through
producers, we could expect that there should be a negative shock impact arising via
exports, and recovery from a financial shock may extend well after the crisis is over15
. If
the impacts occur through consumers, there should be a negative shock on trade flow at
current time when importer had a banking crisis. The negative impacts after importer’s
banking crises were ended also might be shown up in the results. For the similar reason,
there might be significant impact for the time period before crisis.
15
As mentioned in Manova’s (2013)
15
Since a trade might be financed by exporters and importers, both channels may be
important. We also think there might be a bilateral influence occurs there.
1.5.1 One-stage results for exporters and importers
In Table 1.3, the dependent variable is log of the bilateral trade value. The log of
the distance between country pairs is represented by Ln(distance). Contiguity and
common language are the dummy variables for two trade partners who share the same
border or the same official language. GDP Share weighted log-distance, contiguity and
common language is the linear approximation of the inward multilateral resistance and
outward multilateral resistance. The variables Banking crisis for exporter and Banking
crisis for importer are the dummy variables for exporter or importer has a banking crisis
in the current year. The variables N years16
forward of exporter/importer are the dummy
variables for the nth year’s period forward of the beginning year of the variable banking
crisis for exporter/importer, which means these variables are equal to unity if current year
is n years before the beginning of exporter/importer had a banking crisis. The variables N
years lag of exporter/importer are the dummy variables for the nth year’s period lag of
the ending year of the variable banking crisis for exporter/importer, which means these
variables are equal to unity if current year is n years after the ending of exporter/importer
had a banking crisis.
In the first column of Table 1.3, we can see for exporters’ country has a banking
crisis, there is no significant trade flow change, which is not the same as for the channel
16
N years represent one, two three, four and five years corresponding to the variables in the table.
16
through producers. For importers’ country, we can see there is a 21.9% decrease on
average at the current year. This decline is relatively large.
The effect of a banking crisis on trade flow may persist overtime, the second
column of Table 1.3 includes three years time lag to access this impact on trade. For
exporters, there is the lag of banking crisis seem to have no significant impact on export.
For importers, there is a 24.4% decline at the current year of crisis. After crisis was ended,
the trade value tend to decrease even more in the lag time period, it decreased by 30.5%
at the second year after banking crisis ended. After that, the decline become smaller and
back to 25.6% at third lag year.
The third column of Table 1.3 extends the time lag into five years. It shows
almost the same pattern as the second column. For exporter, there are no significant
impacts. For importer, the decline for the current year of crisis is 25.5%, and it still
intensified to the second year after crisis was ended, which is 32.1%. After that, the
contraction tends to recover slowly and back to 24.7% for the fifth year after ending of a
banking crisis. So the sticky effect showed up only for importer.
Column one to three in Table 1.3 provides information about the correlation
between bilateral trade fluctuation and banking crisis for the current year and the time
period after crisis. Trade may change before the onset of a banking crisis. Trade may
increase if there is a bubble in the banking sector or other financial sector which led to a
credit expansion. This expansion can impact the exporter or importer through producers
channel and consumer channel and led to a change in trade flow. Another potential case
17
is that the economy simply was in a recession and the recession caused the both banking
crisis and a decrease in trade value at the same time. Thus information about the trend of
trade flow before banking crisis is required. From fourth column of Table 1.3, for
exporters, there is still no significant influence for all the time, cross through three years
before crisis was begun to three years after crisis was ended. For importers, compare with
the countries don’t have banking crisis, there is average 16%17
decline for the time period
before crisis. This decline trend is relative constant. From these forward years’ results,
the second case that recession already impacted trade flow before crisis seems dominates.
The current year’s value and lag years’ values are almost the same as the second column.
The decline at the current year for crisis is 25.5%. Compare to the year before crisis,
there is an around 10% drop in trade value.
Similarly, the fifth column of Table 1.3 extends both forwards and lags into five
years and get the almost the same information from fourth column of Table 1.3. For
exporter, there is no significant influence on trade flow comes from banking crisis sector
for all the time period. For importer, there is a relative constant 18%18
decline trend.
When banking crisis begun, the decline extents to 27.2%, and it keeps intensified to 33.2%
at the second year after crisis was ended, after that, the decline tends to recover, at fifth
year after banking crisis, the decline is 25.2%
It seems almost all of the influence from banking crisis occurs through the
importers’ channel. For exporters, the financial shock tends to have no significant
17
Low point is 15.5% at two years before crisis. High point is 17.3% at the year right before crisis. 18 Low point is 16.8% at two years before crisis. High point is 18.9% at the year right before crisis.
18
economic impact on bilateral trade flows. For importers, there is constant decline before
banking crisis began. When crisis happened, there is around 10% drop more compare to
the previous decline. The negative impact trend keeps going down another 6% to the
second year after crisis was over, then it starts to recover slowly. From forward years’
results, it seems second scenario: recession before crisis, dominates.
Usually we expect that banking crisis will have a negative impact on bilateral
trade flows, especially focused on the short time period. That’s what we observed in this
great recession and the information from Table 1.3. However, there is another correlation
between banking crisis and financial development. By the fact that the probability of
having a banking crisis is also highly correlated with financial development, the baseline
of the trade value might be higher for a country ever experienced a banking crisis. Some
countries, like North Korea, never have a banking crisis, but also are less developed
financially and trade relatively less.
In Table 1.4, we use the same controls that were used in Table 1.3, and also
include dummy variables that exporter country and importer country ever had at least one
banking crisis cross year 1976 to 2008. We redo the same regressions as in Table 1.3.The
pattern of the banking crisis influence on trade flow on different time period is almost the
same as it was showed in Table 1.3.
For exporters, we see little impacts of a banking crisis on export. For importers,
on the other hand, trade flows appear to decline, on average, leading up to the recession.
For current year had a banking crisis, the negative impact ranges from 21.4% in first
19
column with no lag or forward to 29.7% in fifth column with 5 years forward and lag.
The trade flow suffered a 10% drop compare to the year right before crisis. After crisis
ended, the negative impact intensified to the second lag year, and then recovers slowly.
This pattern is similar to the pattern showed in Table 1.3. Also, the levels of the
coefficients are close to the results from Table 1.3.
From these results in Table 1.4, it seems majority of the influence from financial
shock goes through the importers’ channel, which is robust to the results from Table 1.3.
The persistence of impact for importers is also relatively robust compare to the similar
effect in Table 1.3. Also, Table 1.4 is consistent with the trade impacts on imports prior
to the banking crisis.
For exporters, if the country ever had experienced banking crisis does not seem to
influence bilateral trade flows. For importers, in first column, not includes any forwards
and lags, country ever had a banking crisis tend to trade 2.2% less, however only
significant at 5% level. When we include more information about banking crisis into the
regression, this coefficient tends to be positive and become higher. In third column, when
includes five lag years, a country ever had a banking crisis tends to trade 5.1% higher.
When includes five forward and lag years. The magnitude extends to 10.0%. This result
supports the previous assumption that a country ever experienced a banking crisis is
correlate with higher financial development, thus tends to trade more.
Overall, the results from Table 1.4 reflect that the time pattern for banking crisis
seems to be robust compare to Table 1.3. The countries which ever had experienced at
20
least one banking crisis tend to trade more for importer, but there is no impact for
exporter.
1.5.2 Bilateral results
In Table 1.5, we analyze the bilateral effects of a banking crisis, which captures
the average influence of financial shocks on importers and exporters. The control
variables are the same as previous tables. The variable one crisis ever is a dummy
variable if at least one country of the trade partners has experienced banking crises from
the year 1976-2008. The variable both crises ever equals one if both trade partners have
experienced at least one banking crisis in these years. These two dummy variables are not
mutually exclusive. The dummy variable one crisis equals one if either the importer or
exporter has a banking crisis in the current year, but not if both countries have crises.
Two crises is the dummy variable for if both importer and exporter have banking crises in
the current year. Thus one crisis and two crises are mutually exclusive. The variables N
years lag of one crisis are the dummy variables for the nth year lags of the ending year of
the variable one crisis, which means these variables are equal to unity if current year is n
years after the ending of one county’s banking crisis. The variables N years lag of two
crises are the nth year lags of the ending of two crises, which means these dummy
variables are equal to unity if current year is n years after the ending of the both counties’
banking crises. Since two crises is equal to unity only when both importer and exporter
have banking crises in the current year, there is a case that country A had a banking crisis
before country B’s banking crisis, however it ends during country B’s crisis’ period. In
21
this scenario, two crises is inside of the duration of the one crisis and it will separate the
time period of one crisis into two. However, n years lag of one crisis will not take into
account the ending year of this kind of gap. It only takes account the years after the one
crisis period. When both trade partners are have emerged from their banking crisis, that
year will be treated as the ending of the one crisis, and the lag years will start at that time.
For example in Figure 1.2, Country A had a banking crisis from time t-2 to t+1
and country B had a banking crisis from t to t+3. The overlap time of t to t+1 is
represented by variable two crises. The time period t-2 to t and t+1 to t+3 is represented
as one crisis. The lag year will be t+4, t+5 and so on. The forward year which will be
mentioned later is t-3, t-4 and so on. From the picture, two crises creates a gap inside of
the duration of one crisis, however, the time t won’t be treated as the ending of one crisis
period and t+1 won’t be treated as the beginning of one crisis period.
In the first column of Table 1.5, we can see there is a much higher value of trade
between country pairs if at least one has had a banking crisis. On average, if one trade
partner had banking crises ever, the trade value is around 180.7% higher. If both trade
partners had banking crises ever, the trade value will add another 241.8%. This result is
quite stable when this paper includes banking crisis dummy variables and the time lag
variables.
In the second column of Table 1.5, when one of the trade partners has a banking
crisis, the bilateral trade value is 6.8% higher during the year of the crisis. When both
trade partners have banking crises, the trade volume is 7.6% higher, but this only
22
significant at the 5% level. The signs are different from what we expected. The higher
level of trade in the current year for exporters might be caused by two reasons. One
reason is the regression is using annual data, so the results will capture the average
impact of the whole year. If a banking crisis happened in the second half of the year, then
at least 50% of the current year time was not directly impacted by the banking crisis.
Leaven and Valencia (2012) indicated that August, September and December have higher
frequencies of starting of banking crises as compared to other months. Another reason is
financial crises may impact the trade with a lag. When there is a shock in the financial
intermediary service, manufacture can sustain current production for a time by using
previous savings. The negative impact of the banking crisis will not influence the trade
immediately.
The third column of table 1.5 includes three years time lag of the banking crisis to
test if the financial shock had an impact on trade after it was ended. It shows if one trade
partners suffered a banking crisis, the trade value decreased by around 10%19
each year
after the crisis was over. The current impact of two crises is positive but not significant
for reasons mentioned before. The lag time impact of two crises is negative but also not
quite significant. This means that the presence of a banking crisis in the second country
will not tend to, but not necessarily intensify trade flows decreasing between the two
crises-stricken nations. As discussed previously, since two crises is usually in the gap
between two one crisis, the impact of lag of two banking crises and one banking crisis
might be overlapped.
19
Low point is 9.1% for the second lag year, high point is 12.0% for the first lag year.
23
The fourth column of table 1.5 extends the time lags into five years. It shows
when one of the trade partners suffered a banking crisis, the negative impact on trade
even last for five years after it was ended. The other results are similar to the results in
third column of Table 1.5.
For the lag trend in Table 1.3 and Table 1.4, there is an increasing decline in first
two years after crisis, then recover slowly. The coefficient in Table 1.5 shows the
negative sticky effect is still there, however, the magnitude seems to be relatively
constant.
1.5.3 Two-stage results for exporters and importers
Due to the insignificant results for exporters from previous Tables, this paper uses
two-stage model to test the previous results. In the first stage, Table 1.6 uses the same
control as before. The coefficients for Exporter-year fixed effect and Importer-year fixed
effect capture all the non-bilateral effect for importers and exporters. These unilateral
effects cross time contain the information of outward/inward multilateral resistance,
exporters/importers’ GDP and effect from financial shocks.
Table 1.7 uses coefficients of exporter-year fixed effect from Table 1.6 minus log
of exporters’ GDP as dependent variable to analyze the unilateral effect of banking crisis
on exporters’ side. Importers’ GDP share weighted log distance, common language and
contiguity is the linear approximation of outward multilateral resistance. Compare to the
exporter’s results from Table 1.3 and 1.4, the results in Table 1.7 are largely changed.
24
From column one of Table 1.7, when the exporter had a banking crisis in the
current year, there is a positive 2.7% increase in bilateral trade. When we include the
forward and lag time periods in column 3, the positive impact is around 3.5%. The
second column of Table 1.7 presents a positive impact in the lag time period. There is no
clear trend for this impact. On average, it is around 3%20
. In third column of Table 1.7,
regression includes the forward time period. The lag time period show the similar results
as in second column of Table 1.7. The forward time presents a 3%21
negative impact
from three years before banking crisis to the year right before. For exporters ever had a
banking crisis, the trade value tend to be around 15%22
higher than the courtiers never
experienced a banking crisis.
Overall, the results in Table 1.7 are significant. However, most of the impacts are
economically relatively small. Positive impacts at the current year and lag year are quite
small. The negative impact in the time period before crisis suggests the trade flow
declines prior to the banking crisis, which is consistent to the previous assumption. The
exporter ever had banking crisis used to be insignificant, now tend to be relative large.
Table 1.8 uses coefficients of importer-year fixed effect from Table 4 minus log
of importers’ GDP as dependent variable to analyze the unilateral effect of banking crisis
on importers’ side. In the first column of Table 1.8, there is a 3.7% decline at the current
year when the importer had a banking crisis. The third column of Table 1.8 shows the
20
Low point is 2.2% at five years after crisis. High point is 5.1% at four years after crisis. 21 Low point is 2.2% at one year before crisis. High point is 3.3% at two years before crisis. 22
Low point is 15.1% when include forward and lag time period, high point is 16.1% when not include
forward and lag time period.
25
impacts when we introduce forwards and lags of the banking crisis, we see and 8.0%
decline for the current year. Compare to the results from Table 1.3, Table 1.4, results
from Table 1.8 have same sigh but the magnitude of the decline is smaller.
The second column of Table 1.8 also shows that bilateral trade declines initially
and continues to decline for the two years following the end of a banking crisis. For the
second year lag period, the decline in trade is approximately 9.8%, for the fifth year lag,
the decline in trade is around 3.1%. Compared to the results of column three of Table 1.3
and column three of Table 1.4, the sluggishness of the trade to rebound follows a crisis
continues to hold, however, the magnitude is smaller in this specification.
In third column of Table 1.8, the regression specification includes leads as well as
lags. The lag time period show the similar results as in second column of Table 1.8. The
forward time period also exhibits a constant 14%23
negative trend, which is similar to the
forward trends in fifth column of Table 1.3 and fifth column of Table 1.4 but with a
smaller level. In Table 1.3 and 1.4, there is decline in bilateral trade value when the time
moves from the year before crisis to the crisis year. However, we cannot observe this
pattern in Table 1.8.
For importers, the coefficients of dummy variable for “ever had a banking crisis”
differs from the previous specifications. In the third column of Table 1.8, which includes
forwards and lags, countries that ever had a banking crisis tends to trade 2.9% less. This
23
Low point is 12.9% at four years before crisis. High point is 15.5% at three years before crisis.
26
is different from the positive results from Table 1.4 and bilateral effects from Table 1.5. It
is also against with our previous assumption.
Overall, results and most of the pattern for different time period of banking crisis
from Table 1.8 are similar to those in Table 1.3 and Table 1.4.
1.6 Conclusion
After the Great Recession, the researches focused on how the financial sector
impacts the bilateral trade and the magnitudes of these impacts. This analysis uses the
data that cover most of the countries involved in global trade across 32 years, and
produces relatively robust results, what is the correlation between banking crisis and
fluctuation in trade on average level.
One surprising result is almost all the influence from financial sector for trade
goes to importer’s side, even majority of the theoretical research are focus on the
exporter’s side. In two-stage analysis, there are some significant results for exporter.
However, almost all the effects are economically small24
.
On other side, there is a robust correlation between banking crisis and bilateral
imports. If trade contains a large amount of durable goods and capital goods, trade may
be more sensitive to income. Also almost all the impacts for the different time periods,
especially for the current year of banking crisis and the years after it, are significantly
negative.
24
Average is around 3%.
27
Combining two different influences from exporter and importer, there is one
potential explanation on the 2007 Great Recession in trade flow collapse. Previous
financial shocks, for countries with various income levels, occurred in different time
periods. The overall global market on demand side was only slightly disturbed leaving
the trade flow on stable average level. The Great Recession is wildly spread over the
world and impacted a lot of high income countries, which are also high demand
countries in global trade system. The large drop in demand causes the collapse in trade.
This paper also shows that there is a pattern for the impact of financial shock on
the trade flow for importers. Before the importer has a banking crisis, import is already
in a constant decline compare to other countries. It might be caused by the importer was
already in a recession before the onset of the banking crisis. When crisis begins, the
decline in import extends. Even after banking crisis is over, on average, the decline will
still be intensified to the second year after crisis, and then tends to recover slowly.
Most of the results support that, on average level, countries that had ever
experienced banking crisis tend to trade more. For bilateral effect analysis, this effect is
even larger. These results support the assumption that banking crisis is also correlated
with financial development, and countries with higher financial development tend to
trade more. In future research, the measurement of finical development could be used to
test this assumption.
28
Appendices
29
Appendix A Robustness check for bilateral results
Table 1.5 provides information about bilateral effects for the time period with
banking crisis and after banking crisis. Tables 1.9 include the time period before the
banking crisis.
From the first column of table 1.9, the coefficient for “one country had crisis” is
positive and significant. The coefficient for both countries had crises for the current year
is still positive but not significant. For the country pair that at least one has ever had a
banking crisis, the trade value is about 182.6% higher. If both trade partners had banking
crisis, the trade value will add another 244.5%. At the time when one of the country pair
had a banking crisis the trade value is 5.5% higher. When both countries had banking
crisis, the coefficient is positive but not significant. For the lag time periods, the trade
value is about 10% lower when one country had suffered a banking crisis. However, for
both countries had suffered banking crises, the negative impacts after crisis was over
seem won’t be intensified. The sticky effect is relatively constant. All these results are
quite close to the results from third column of Table 1.5.
In the first column of Table 1.9, estimates provide no support for a clear trend in
trade values in the year prior to a single country’s banking crisis. For the time period
before both country had banking crises, the trade value is around 10% lower, but
relatively not quite significant. In the second column of Table 1.9, the forward and lag
time periods are extend to five years. Almost all the conclusions are remained the same.
30
Even the magnitudes are quite similar to the first column of Table 1.9, which are also
similar to the results from fourth column from Table 1.5.
Overall, the bilateral effects are robust compare to the results from Table 1.5.
There is a relatively constant decline after banking crisis. When one side of the country
pair suffered a banking crisis, a banking crisis occurs on another side will not tend to
intensify the fluctuation of the trade flow between these two countries. For the years
ahead of banking crisis, there is no clear trend.
31
Appendix B Robustness check for Two-stage results
Table 1.10 replicates the regressions in Table 1.5 and include importer-exporter
fixed effect which will capture all the time invariant bilateral effects. The bilateral
banking crisis effects from Table 1.10 are consistent to the results from Table 1.5
The coefficients from Table 1.11 show that when one country of the trade pair
had ever experienced banking crisis, the trade flow tends to increased by around 65%. If
both countries of the trade pair had ever experienced banking crises, the trade flow tends
to increased by another 120%. These results are consistent to what we found in Table 1.5.
Compare to the results in Table 1.5, the magnitude is relatively smaller. However, they
are still higher than the corresponding results from unilateral effect estimations.
The results for exporter and importer from Table 1.12 to 1.13 seem to be puzzle.
The signs changed when include exporter/importer ever had experienced a banking crisis.
It might be correlated with the restriction that coefficient of log GDP is equal to -1 in the
dependent variables.
32
Afghanistan Ghana Pakistan
Albania Greece Panama
Algeria Greenland Papua New Guinea
Angola Guadeloupe Paraguay
Argentina Guatemala Peru
Armenia Guinea Philippines
Aruba Guinea-Bissau Poland
Australia Guyana Portugal
Austria Haiti Qatar
Azerbaijan Honduras Reunion
Bahamas Hong Kong Romania
Bahrain Hungary Russian Federation
Bangladesh Iceland Rwanda
Barbados India Saint Kitts and Nevis
Belarus Indonesia Samoa
Belgium and Luxembourg Iran Saudi Arabia
Belize Iraq Senegal
Benin Ireland Serbia and Montenegro
Bermuda Israel Seychelles
Bhutan Italy Sierra Leone
Bolivia Jamaica Singapore
Bosnia and Herzegovina Japan Slovakia
Brazil Jordan Slovenia
Bulgaria Kazakstan Somalia
Burkina Faso Kenya South Africa
Burma Kiribati Spain
Burundi Korea Sri Lanka
Cambodia Kuwait Sudan
Cameroon Kyrgyzstan Suriname
Canada Lao People's Democratic Republic Sweden
Central African Republic Latvia Switzerland
Chad Lebanon Syrian Arab Republic
Chile Liberia Taiwan
China Libyan Arab Jamahiriya Tajikistan
Colombia Lithuania Tanzania, United Rep. of
Comoros Macau (Aomen) Thailand
Congo Macedonia (the former Yugoslav Rep. of) Togo
Congo (Democratic Republic of the) Madagascar Trinidad and Tobago
Costa Rica Malawi Tunisia
Croatia Malaysia Turkey
Cyprus Mali Turkmenistan
Czech Republic Malta Uganda
Côte d'Ivoire Mauritania Ukraine
Denmark Mauritius United Arab Emirates
Djibouti Mexico United Kingdom
Dominican Republic Micronesia (Federated States of) United States of America
Ecuador Moldova, Rep.of Uruguay
Egypt Mongolia Uzbekistan
El Salvador Morocco Venezuela
Equatorial Guinea Mozambique Viet Nam
Estonia Nepal Yemen
Ethiopia Netherland Antilles Zambia
Fiji Netherlands Zimbabwe
Finland New Caledonia
France New Zealand
French Guiana Nicaragua
Gabon Niger
Gambia Nigeria
Georgia Norway
Germany Oman
Table 1.1: Names of Countries and Districts
33
No of obs. Mean Std. dev. Min. Max
Trade (1000 current US dollar) 420960 3.490e+5 3.376e+6 0.001 3.31e+8
Country’s GDP (exporter, 1000 current US dollar) 847008 1.67e+8 7.53e+8 2.057e+4 1.42e+10
Distance 987657 7653.3 4429.0 1.881 1.995e+4
Contiguity 987657 0.019 0.135 0 1
Common language 987657 0.137 0.344 0 1
Banking crisis (exporter) 987657 0.065 0.246 0 1
One country of the trade pair has banking crisis 987657 0.116 0.320 0 1
Both countries of the trade pair have banking crises 987657 0.007 0.084 0 1
One country of the trade pair ever had banking crisis 987657 0.863 0.344 0 1
Both countries of the trade pair ever had banking crises 987657 0.397 0.489 0 1
Table 1.2: Summary Statistics for Chapter 1
34
Dep var ln(trade)
Ln(distance) minus importers’ share weighted log distance
minus exporters’ share weighted log distance
-1.109***
(0.006)
-1.110***
(0.006)
-1.110***
(0.006)
-1.109***
(0.006)
-1.108***
(0.006)
Contiguity minus importers’ share weighted contiguity
minus exporters’ share weighted contiguity
0.363***
(0.025)
0.373***
(0.025)
0.379***
(0.025)
0.377***
(0.025)
0.386***
(0.025)
Common language minus importers’ share weighted
language minus exporters’ share weighted language
0.572***
(0.011)
0.577***
(0.011)
0.581***
(0.011)
0.581***
(0.011)
0.586***
(0.011)
Ln(exporter ’s GDP) 1.226***
(0.002)
1.227***
(0.002)
1.227***
(0.002)
1.227***
(0.002)
1.227***
(0.002)
Ln(importer’s GDP) 1.081***
(0.002)
1.082***
(0.002)
1.082***
(0.002)
1.084***
(0.002)
1.086***
(0.002)
5 years forward of banking crisis for exporter 0.023
(0.027)
4 years forward of banking crisis for exporter -0.008
(0.026)
3 years forward of banking crisis for exporter -0.020
(0.026)
-0.021
(0.026)
2 years forward of banking crisis for exporter -0.022
(0.025)
-0.022
(0.025)
1 years forward of banking crisis for exporter -0.024
(0.025)
-0.024
(0.024)
Banking crisis for exporter 0.005
(0.015)
0.004
(0.015)
0.002
(0.015)
0.002
(0.015)
0.001
(0.015)
1 years lag of banking crisis for exporter -0.010
(0.025)
-0.011
(0.025)
-0.011
(0.025)
-0.012
(0.025)
2 years lag of banking crisis for exporter -0.020 (0.024)
-0.021 (0.025)
-0.021 (0.024)
-0.023 (0.025)
3 years lag of banking crisis for exporter 0.001
(0.025)
-0.002
(0.024)
0.000
(0.024)
-0.003
(0.024)
4 years lag of banking crisis for exporter -0.013 (0.024)
-0.015 (0.024)
5 years lag of banking crisis for exporter -0.034
(0.024)
-0.035
(0.024)
5 years forward of banking crisis for importer -0.199*** (0.026)
4 years forward of banking crisis for importer -0.186***
(0.026)
3 years forward of banking crisis for importer -0.176*** (0.026)
-0.196*** (0.026)
2 years forward of banking crisis for importer -0.168***
(0.025)
-0.185***
(0.025)
1 years forward of banking crisis for importer -0.190*** (0.024)
-0.210*** (0.024)
Banking crisis for importer -0.247***
(0.015)
-0.280***
(0.015)
-0.294***
(0.015)
-0.295***
(0.015)
-0.317***
(0.015)
1 years lag of banking crisis for importer -0.321*** (0.026)
-0.337*** (0.025)
-0.334*** (0.026)
-0.357*** (0.026)
2 years lag of banking crisis for importer -0.364***
(0.025)
-0.387***
(0.025)
-0.374***
(0.025)
-0.403***
(0.025)
3 years lag of banking crisis for importer -0.296*** (0.025)
-0.319*** (0.024)
0.305*** (0.025)
-0.332*** (0.024)
4 years lag of banking crisis for importer -0.289***
(0.024)
-0.301***
(0.024)
5 years lag of banking crisis for importer -0.284***
(0.024)
-0.290***
(0.023)
Constant -40.292***
(0.083)
-40.294***
(0.083)
-40.300***
(0.083)
-40.322***
(0.083)
-40.511***
(0.083)
Year fixed effect Yes Yes Yes Yes Yes
R-square 0.635 0.636 0.636 0.636 0.636
No of obs. 379928 379928 379928 379928 379928
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.3: Linear approximations for multilateral resistance and banking crises
35
Dep var ln(trade)
Ln(distance) minus importers’ share weighted log distance
minus exporters’ share weighted log distance
-1.108***
(0.006)
-1.110***
(0.006)
-1.111***
(0.006)
-1.110***
(0.006)
-1.111***
(0.006)
Contiguity minus importers’ share weighted contiguity
minus exporters’ share weighted contiguity
0.367***
(0.025)
0.371***
(0.025)
0.375***
(0.025)
0.373***
(0.025)
0.378***
(0.025)
Common language minus importers’ share weighted
language minus exporters’ share weighted language
0.575***
(0.011)
0.576***
(0.011)
0.577***
(0.011)
0.577***
(0.011)
0.579***
(0.011)
Ln(exporter ’s GDP) 1.227***
(0.002)
1.227***
(0.002)
1.227***
(0.002)
1.227***
(0.002)
1.228***
(0.002)
Ln(importer’s GDP) 1.082***
(0.002)
1.081***
(0.002)
1.079***
(0.002)
1.082***
(0.002)
1.082***
(0.002)
5 years forward of banking crisis for exporter 0.024
(0.027)
4 years forward of banking crisis for exporter -0.007
(0.026)
3 years forward of banking crisis for exporter -0.018
(0.026)
-0.020
(0.026)
2 years forward of banking crisis for exporter -0.019
(0.025)
-0.021
(0.025)
1 years forward of banking crisis for exporter -0.022
(0.025)
-0.023
(0.024)
Banking crisis for exporter 0.008
(0.015)
0.007
(0.015)
0.005
(0.015)
0.004
(0.015)
0.002
(0.015)
1 years lag of banking crisis for exporter -0.006
(0.025)
-0.008
(0.025)
-0.008
(0.025)
-0.010
(0.025)
2 years lag of banking crisis for exporter -0.016 (0.024)
-0.019 (0.025)
-0.018 (0.024)
-0.021 (0.025)
3 years lag of banking crisis for exporter 0.004
(0.025)
0.001
(0.024)
0.003
(0.024)
-0.001
(0.024)
4 years lag of banking crisis for exporter -0.011 (0.024)
-0.013 (0.024)
5 years lag of banking crisis for exporter -0.031
(0.024)
-0.033
(0.024)
5 years forward of banking crisis for importer -0.228*** (0.027)
4 years forward of banking crisis for importer -0.217***
(0.026)
3 years forward of banking crisis for importer -0.188*** (0.026)
-0.228*** (0.026)
2 years forward of banking crisis for importer -0.181***
(0.025)
-0.216***
(0.025)
1 years forward of banking crisis for importer -0.203*** (0.024)
-0.243*** (0.024)
Banking crisis for importer -0.241***
(0.015)
-0.286***
(0.015)
-0.311***
(0.015)
-0.310***
(0.015)
-0.352***
(0.015)
1 years lag of banking crisis for importer -0.328*** (0.025)
-0.355*** (0.025)
-0.349*** (0.026)
-0.394*** (0.026)
2 years lag of banking crisis for importer -0.370***
(0.025)
-0.405***
(0.025)
-0.389***
(0.025)
-0.440***
(0.025)
3 years lag of banking crisis for importer -0.302*** (0.024)
-0.337*** (0.024)
0.320*** (0.024)
-0.368*** (0.024)
4 years lag of banking crisis for importer -0.306***
(0.024)
-0.335***
(0.024)
5 years lag of banking crisis for importer -0.301***
(0.024)
-0.323***
(0.023)
Exporter ever had a banking crisis -0.010
(0.009)
-0.009
(0.009)
-0.007
(0.009)
-0.007
(0.009)
-0.005
(0.009)
Importer ever had a banking crisis -0.022* (0.009)
-0.020* (0.009)
0.050*** (0.009)
0.043*** (0.009)
0.095*** (0.010)
Constant -40.287***
(0.083)
-40.295***
(0.083)
-40.307***
(0.083)
-40.328***
(0.083)
-40.529***
(0.083)
Year fixed effect Yes Yes Yes Yes Yes
R-square 0.635 0.636 0.636 0.636 0.636
No of obs. 379928 379928 379928 379928 379928
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.4: Linear approximations for multilateral resistance and banking crises with
country ever experienced a banking crisis
36
Dep var. ln(trade)
Ln(distance) -1.382***
(0.004)
-1.382***
(0.004)
-1.382***
(0.004)
-1.383***
(0.004)
Contiguity 0.520***
(0.021)
0.521***
(0.021)
0.520***
(0.021)
0.520***
(0.021)
Common language 0.876***
(0.009)
0.876***
(0.009)
0.876***
(0.009)
0.876***
(0.009)
One crisis ever 1.032***
(0.011)
1.025***
(0.011)
1.039***
(0.011)
1.048***
(0.011)
Both crises ever 1.229***
(0.006)
1.223***
(0.007)
1.234***
(0.007)
1.242***
(0.007)
One crisis 0.066***
(0.009)
0.051***
(0.010)
0.044***
(0.009)
One year lag of one
crisis
-0.128***
(0.016)
-0.127***
(0.016)
Two years lag of
one crisis
-0.095***
(0.015)
-0.095***
(0.016)
Three years lag of
one crisis
-0.112***
(0.015)
-0.117***
(0.015)
Four years lag of
one crisis
-0.092***
(0.015)
Five years lag of
one crisis
-0.135***
(0.015)
Two crises 0.073*
(0.036)
0.049
(0.035)
0.035
(0.035)
One year lag of two crises
-0.111* (0.050)
-0.107* (0.050)
Two years lag of
two crises
-0.062
(0.049)
-0.056
(0.049)
Three years lag of two crises
-0.115* (0.048)
-0.112* (0.048)
Four years lag of
two crises
-0.046
(0.047)
Five years lag of two crises
-0.107* (0.046)
Importer-year
fixed effect
yes yes yes yes
Exporter-year fixed effect
yes yes yes yes
Constant 17.946***
(0.037)
17.947***
(0.037)
17.947***
(0.037)
17.947***
(0.037)
R-square 0.74 0.74 0.74 0.74
No of obs. 420960 420960 420960 420960
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.5: Trade value and banking crises with lags
37
Dep. Var. ln(trade)
Ln(distance) -1.386***
(0.004)
Contiguity 0.515***
(0.021)
Common language 0.882***
(0.009)
Importer-year
fixed effect
Yes
Exporter-year
fixed effect
Yes
Constant 19.498***
(0.035)
R-square 0.74
No of obs. 420960
*** for p-value<0.001 ** for p-value<0.01
* for p-value<0.05
Table 1.6: First stage of the regression
38
Coefficient of Exporter-year fixed effect from Table 1.4 minus log of
Exporters’ GDP
Five year forward of exporter’s crisis
0.058*** (0.008)
Four year forward of
exporter’s crisis
0.051***
(0.008)
Three year forward of exporter’s crisis
-0.024** (0.008)
Two year forward of
exporter’s crisis
-0.034***
(0.007)
One year forward of exporter’s crisis
-0.022*** (0.007)
Banking crises for
exporter
0.027***
(0.005)
0.035***
(0.005)
0.034***
(0.005)
One year lag of exporter’s crisis
0.024** (0.008)
0.024** (0.008)
Two year lag of
exporter’s crisis
0.024**
(0.008)
0.024**
(0.008)
Three year lag of
exporter’s crisis
0.047***
(0.008)
0.045***
(0.008)
Four year lag of
exporter’s crisis
0.050***
(0.008)
0.049***
(0.008)
Five year lag of exporter’s crisis
0.022** (0.008)
0.022** (0.008)
Exporter ever had a
banking crisis
0.149***
(0.003)
0.141***
(0.003)
0.141***
(0.003)
Constant -20.075***
(0.035)
-19.958***
(0.034)
-20.606***
(0.035)
Importers’ GDP share
weighted log distance
Yes Yes Yes
Importers’ GDP share
weighted language
Yes Yes Yes
Importers’ GDP share
weighted contiguity
Yes Yes Yes
Year fixed effect Yes Yes Yes
R-square 0.131 0.131 0.131
No of obs 823649 823649 823649
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.7: Exporter-year fixed effect and exporters’ banking crisis
39
Coefficient of Importer-year fixed effect from Table 1.4 minus log of
Importers’ GDP
Five year forward of importer’s crisis
-0.148*** (0.007)
Four year forward of
importer’s crisis
-0.138***
(0.007)
Three year forward of importer’s crisis
-0.168*** (0.007)
Two year forward of
importer’s crisis
-0.149***
(0.007)
One year forward of importer’s crisis
-0.148*** (0.007)
Banking crises for
impoter
-0.038***
(0.004)
-0.053***
(0.004)
-0.083***
(0.004)
One year lag of importer’s crisis
-0.077*** (0.007)
-0.106*** (0.007)
Two year lag of
importer’s crisis
-0.103***
(0.007)
-0.127***
(0.007)
Three year lag of
importer’s crisis
-0.060***
(0.007)
-0.081***
(0.007)
Four year lag of
importer’s crisis
-0.040***
(0.007)
-0.061***
(0.007)
Five year lag of importer’s crisis
-0.032*** (0.007)
-0.049*** (0.007)
Importer ever had a
banking crisis
-0.075***
(0.003)
-0.062***
(0.003)
-0.029***
(0.003)
Constant -17.050***
(0.031)
-17.122***
(0.031)
-17.182***
(0.031)
Exporters’ GDP share
weighted distance
Yes Yes Yes
Exporters’ GDP share
weighted language
Yes Yes Yes
Exporters’ GDP share
weighted contiguity
Yes Yes Yes
Year fixed effect Yes Yes Yes
R-square 0.156 0.156 0.159
No of obs 638716 638716 638716
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.8: Importer-year fixed effect and importers’ banking crisis
40
Dep var ln(trade)
Ln(distance) -1.382***
(0.004)
-1.383***
(0.004)
Contiguity 0.521***
(0.021)
0.521***
(0.021)
Common language 0.876***
(0.009)
0.876***
(0.009)
One crisis ever 1.039***
(0.011)
1.043***
(0.011)
Both crises ever 1.237***
(0.007)
1.244***
(0.007)
Five years forward of one crisis 0.086***
(0.016)
Four years forward of one crisis 0.028
(0.016)
Three years forward of one crisis -0.030
(0.015)
-0.024
(0.016)
Two years forward of one crisis 0.056***
(0.015)
0.057***
(0.015)
One year forward of one crisis -0.040**
(0.015)
-0.040**
(0.015)
One crisis 0.054***
(0.010)
0.049***
(0.010)
One year lag of one crisis -0.124***
(0.016)
-0.123***
(0.016)
Two years lag of one crisis -0.094*** (0.015)
-0.093*** (0.016)
Three years lag of one crisis -0.112***
(0.015)
-0.115***
(0.015)
Four years lag of one crisis -0.091*** (0.015)
Five years lag of one crisis -0.134***
(0.015)
Five years forward of two crises -0.126* (0.052)
Four years forward of two crises -0.138**
(0.047)
Three years forward of two crises -0.153** (0.051)
-0.158** (0.051)
Two years forward of two crises -0.061
(0.048)
-0.065
(0.048)
One year forward of two crises -0.121* (0.047)
-0.122** (0.047)
Two crises 0.046
(0.035)
0.037
(0.035)
One year lag of two crises -0.107* (0.050)
-0.104* (0.050)
Two years lag of two crises -0.060
(0.049)
-0.054
(0.049)
Three years lag of two crises -0.114* (0.048)
-0.111* (0.048)
Four years lag of two crises -0.046
(0.047)
Five years lag of two crises -0.107*
(0.046)
Importer year fixed effect yes yes
Exporter year fixed effect yes yes
Constant 17.951*** (0.037)
17.957*** (0.037)
R-square 0.74 0.74
No of obs. 420960 420960
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.9: Trade value and banking crises with forwards and lags
41
Dep var ln(trade)
Five years forward
of one crisis
0.026*
(0.011)
Four years forward
of one crisis
0.051***
(0.011)
Three years forward
of one crisis
-0.046***
(0.010)
-0.056***
(0.011)
Two years forward
of one crisis
0.036***
(0.010)
0.026*
(0.010)
One year forward of
one crisis
-0.019
(0.010)
-0.028**
(0.010)
One crisis 0.063***
(0.006)
0.052***
(0.006)
0.046***
(0.006)
0.057***
(0.006)
0.058***
(0.006)
One year lag of one
crisis
-0.056***
(0.011)
-0.064***
(0.011)
-0.062***
(0.011)
-0.070***
(0.011)
Two years lag of
one crisis
-0.064***
(0.010)
-0.073***
(0.010)
-0.068***
(0.010)
-0.076***
(0.011)
Three years lag of
one crisis
-0.069***
(0.010)
-0.081***
(0.010)
-0.070***
(0.010)
-0.083***
(0.010)
Four years lag of
one crisis
-0.057***
(0.010)
-0.058***
(0.010)
Five years lag of
one crisis
-0.113***
(0.010)
-0.113***
(0.010)
Five years forward
of two crises
0.092**
(0.035)
Four years forward of two crises
0.169*** (0.035)
Three years forward
of two crises
0.076*
(0.035)
0.084*
(0.035)
Two years forward of two crises
0.155*** (0.033)
0.162*** (0.033)
One year forward of
two crises
0.125***
(0.032)
0.131***
(0.032)
Two crises 0.144*** (0.024)
0.134*** (0.024)
0.125*** (0.024)
0.170*** (0.024)
0.186*** (0.024)
One year lag of two
crises
-0.030
(0.034)
-0.031
(0.034)
-0.019
(0.034)
-0.011
(0.034)
Two years lag of two crises
0.012 (0.033)
0.012 (0.033)
0.022 (0.033)
0.033 (0.033)
Three years lag of
two crises
-0.034
(0.032)
-0.036
(0.032)
-0.023
(0.032)
-0.015
(0.032)
Four years lag of two crises
-0.008 (0.032)
0.013 (0.032)
Five years lag of
two crises
-0.085**
(0.032)
-0.064**
(0.032)
Importer year fixed effect
yes yes yes yes yes
Exporter year
fixed effect
yes yes yes yes yes
Importer-Exporter fixed effect
yes yes yes yes yes
Constant 7.668***
(0.002)
7.677***
(0.002)
7.688***
(0.003)
7.677***
(0.003)
7.682***
(0.003)
R-square 0.878 0.878 0.878 0.878 0.878
No of obs. 420960 420960 420960 420960 420960
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.10: Trade value and banking crises with Importer-Exporter fixed effect
42
Coefficient of
importer-Exporter fixed
effect from
First column of Table 1.8
Coefficient of
importer-Exporter fixed
effect from
Third column of Table 1.8
Coefficient of
importer-Exporter fixed
effect from
Fifth column of Table 1.8
Ln(distance) -1.158***
(0.004)
-1.158***
(0.004)
-1.160***
(0.004)
Contiguity 1.720*** (0.025)
1.722*** (0.025)
1.711*** (0.025)
Common
language
0.131***
(0.009)
0.131***
(0.009)
0.136***
(0.009)
One crisis ever
0.513*** (0.011)
0.520*** (0.011)
0.492*** (0.011)
Both crises
ever
0.823***
(0.007)
0.827***
(0.007)
0.792***
(0.007)
Constant 7.843*** (0.041)
7.829*** (0.041)
7.891*** (0.041)
R-square 0.139 0.139 0.138
No of obs. 846813 846813 846813
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.11: Importer-Exporter fixed effect and time invariant bilateral variables
43
Coefficient
of Exporter-
year fixed
effect from First
column of
Table 1.8 minus log
of
Exporters’ GDP
Coefficient
of Exporter-
year fixed
effect from First
column of
Table 1.8 minus log
of
Exporters’ GDP
Coefficient
of Exporter-
year fixed
effect from Third
column of
Table 1.8 minus log
of
Exporters’ GDP
Coefficient
of Exporter-
year fixed
effect from Third
column of
Table 1.8 minus log
of
Exporters’ GDP
Coefficient
of Exporter-
year fixed
effect from Fifth
column of
Table 1.8 minus log
of
Exporters’ GDP
Coefficient
of Exporter-
year fixed
effect from Fifth
column of
Table 1.8 minus log
of
Exporters’ GDP
Five year forward of
exporter’s crisis
-0.364***
(0.011)
-0.179***
(0.011)
Four year forward of exporter’s crisis
-0.406*** (0.011)
-0.216*** (0.011)
Three year forward of
exporter’s crisis
-0.321***
(0.010)
-0.124***
(0.010)
Two year forward of
exporter’s crisis
-0.410***
(0.010)
-0.203***
(0.010)
One year forward of
exporter’s crisis
-0.389***
(0.010)
-0.177***
(0.010)
Banking crises orientation
-0.278*** (0.006)
-0.092*** (0.006)
-0.292*** (0.006)
-0.063*** (0.006)
-0.347*** (0.010)
-0.118*** (0.006)
One year lag of
exporter’s crisis
-0.091***
(0.011)
0.145***
(0.011)
-0.111***
(0.011)
0.118***
(0.011)
Two year lag of exporter’s crisis
-0.086*** (0.011)
0.155*** (0.011)
-0.096*** (0.011)
0.131*** (0.011)
Three year lag of
exporter’s crisis
-0.061***
(0.011)
0.176***
(0011)
-0.063***
(0.011)
0.155***
(0.011)
Four year lag of exporter’s crisis
-0.080*** (0.011)
0.153*** (0.011)
-0.076*** (0.011)
0.135*** (0.011)
Five year lag of
exporter’s crisis
-0.066***
(0.011)
0.162***
(0.011)
-0.059***
(0.011)
0.145***
(0.011)
Exporter ever had a banking crisis
-0.527*** (0.004)
-0.552*** (0.004)
-0.482*** (0.004)
Constant -25.161***
(0.047)
-23.656***
(0.048)
-25.156***
(0.047)
-23.466***
(0.047)
-24.960***
(0.047)
-23.555***
(0.047)
Importers’ GDP share weighted distance
Yes Yes Yes Yes Yes Yes
Importers’ GDP share
weighted language
Yes Yes Yes Yes Yes Yes
Importers’ GDP share weighted contiguity
Yes Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes Yes
R-square 0.256 0.274 0.256 0.274 0.260 0.272
No of obs 832649 832649 832649 832649 832649 832649
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.12: Exporter-year fixed effect and exporters’ banking crisis for robustness check
44
Coefficient
of Importer-
year fixed
effect from First
column of
Table 1.8 minus log
of
Importers’ GDP
Coefficient
of Importer-
year fixed
effect from First
column of
Table 1.8 minus log
of
Importers’ GDP
Coefficient
of Importer-
year fixed
effect from Third
column of
Table 1.8 minus log
of
Importers’ GDP
Coefficient
of Importer-
year fixed
effect from Third
column of
Table 1.8 minus log
of
Importers’ GDP
Coefficient
of Importer-
year fixed
effect from Fifth
column of
Table 1.8 minus log
of
Importers’ GDP
Coefficient
of Importer-
year fixed
effect from Fifth
column of
Table 1.8 minus log
of
Importers’ GDP
Five year forward of
importer’s crisis
-0.560***
(0.013)
-0.333***
(0.012)
Four year forward of importer’s crisis
-0.650*** (0.013)
-0.419*** (0.012)
Three year forward of
importer’s crisis
-0.645***
(0.013)
-0.413***
(0.013)
Two year forward of
importer’s crisis
-0.608***
(0.012)
-0.369***
(0.012)
One year forward of
importer’s crisis
-0.542***
(0.012)
-0.292***
(0.012)
Banking crises destination
-0.435*** (0.007)
-0.202*** (0.007)
-0.468*** (0.008)
-0.187*** (0.007)
-0.541*** (0.007)
-0.268*** (0.008)
One year lag of
importer’s crisis
-0.250***
(0.013)
0.036**
(0.012)
-0.293***
(0.012)
-0.019
(0.012)
Two year lag of importer’s crisis
-0.241*** (0.012)
0.053*** (0.012)
-0.269*** (0.012)
0.005 (0.012)
Three year lag of
importer’s crisis
-0.159***
(0.012)
0.127***
(0.012)
-0.180***
(0.012)
0.083***
(0.012)
Four year lag of importer’s crisis
-0.184*** (0.012)
0.096*** (0.012)
-0.203*** (0.012)
0.053*** (0.012)
Five year lag of
importer’s crisis
-0.100***
(0.012)
0.180***
(0.012)
-0.110***
(0.012)
0.142***
(0.012)
Importer ever had a banking crisis
-0.766*** (0.004)
-0.777*** (0.005)
-0.675*** (0.005)
Constant -24.304***
(0.051)
-22.422***
(0.053)
-24.449***
(0.053)
-22.362***
(0.052)
-24.297***
(0.052)
-22.529***
(0.052)
Exporters’ GDP share weighted distance
Yes Yes Yes Yes Yes Yes
Exporters’ GDP share
weighted language
Yes Yes Yes Yes Yes Yes
Exporters’ GDP share weighted contiguity
Yes Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes Yes
R-square 0.253 0.285 0.254 0.285 0.265 0.286
No of obs 638716 638716 638716 638716 638716 638716
*** for p-value<0.001 ** for p-value<0.01 * for p-value<0.05
Table 1.13: Importer-year fixed effect and importers’ banking crisis for
robustness check
45
Figure 1.1: Import goods and services as a ratio of GDP for a select group of
countries with banking crises at duration of two years
0.2
.4.6
0 5 10 15var1
BGR CAF
CRI DOM
GHA
.2.4
.6.8
1
0 5 10 15var1
HRV KOR
LTU LVA
NIC
.2.4
.6.8
1
0 5 10 15var1
PAN SLV
TGO TUR
TZA UKR
46
Figure 1.2: Interpretation of forward and lag time
47
Chapter 2
Banking Crises and the Impacts on the
Margins of Trade
2.1 Introduction
Since 2007 banking crisis and the onset of Great Recession, the investigation of
the collapse of bilateral trade has evolved into a cottage industry among trade economists.
Most of literatures provide a link between the Great Recession and trade fall in
international trade focusing on this event. In Chapter 1 of this dissertation, we focused on
how banking crisis may influence the bilateral trade flows over time. It attempted to
disentangle the financial shock’s impact on trade flows that seemingly originated on the
export side and those that originated on the import side.
Based on the research on financial shocks for both bilateral effect and unilateral
effect, the objective of this paper is to assess how the financial shocks impact the
extensive and intensive margins of trade. To this end, we decompose bilateral trade flows
into two parts: extensive margin, which reflects the information on a share weighted
48
count of the number of varieties of goods traded1, and the intensive margin, which
presents the volume of each variety. As a result of this decomposition, we can assess how
financial crises impact the number of goods shipped and volume of goods shipped.
Theoretically, it is not clear how the crisis will impact these margins. Hence, “Whether
there is a robust correlation between banking crises and extensive/intensive margins
fluctuations?” will be the main topic of this paper
As in Chapter 1, financial shocks may influence the extensive and intensive
margin through two channels. One occurs through the producers: when trade is financed
by exporters, a tightened credit constraint may force some producers to exit foreign
markets and cause a decline on extensive margin for export. The rising cost for financing
the trade will also have an impact on intensive margin from export side. The other
channel occurs through consumers when trade is financed by importers. There might be a
large decline in capital goods and durable goods, which are relatively income sensitive. It
might have a negative impact on both extensive margin and intensive margin for import.
However, when there is a decline on extensive margin and hence import less varieties of
goods caused by income effect, there might be a substitution effect and caused the change
in value of import for each variety and change the intensive margin in a positive direction.
In this paper, we will use Hummels and Klenow (2005)’s method to decompose the
margins and try to uncover the average impact for exporters and importers on both
extensive and intensive margin.
1 This paper defines variety at industry level.
49
In Chapter 1, we found that there was an impact for importers in advance of the
banking crises, and recovery extended well after the crisis was over. In order to capture
the timing of the effects and to be consistent with the modeling strategy in that
chapter ,we also include different time periods to find out whether there is any patterns
for impacts from banking crisis cross time on different margins.
2.2 Literature Review
Hummels and Klenow (2005) provide a method to decompose the bilateral trade
flow into extensive and intensive margin. They show that higher income countries tend to
export more varieties of goods. Their paper also shows that the majority of the bilateral
trade can be attributed to extensive margin. By adopting their method, the trading cost
can be estimated individually for both margins. Bernard et al. (2007) use U.S. firm level
data, research the distance effect on extensive and intensive margin.
The literatures on the trade collapse during Great Recession highlights factors that
contributed to our understanding of factors that are associated with the decline in trade
flows during the financial crisis. Levchenko et al. (2010) uses disaggregated quarterly US
trade data and finds a great decline in the volume of trade. Chor and Manova (2012) used
monthly US import data to analysis the trade collapse after the 2007 crisis, finding that
the exports of industries with larger dependence on the external financial market will tend
to be more vulnerable and sensitive to financial shocks. Lacovone and Zavacka (2009)
use annual data to show that the industries which are more dependent on financial
markets in more financially developed countries experienced larger declines in trade
50
during the banking crisis. Bricongne et al. (2012) uses French firm level data, found out
extensive margin and financial constraints played a minor role in the French export.
There are also studies on trade collapse and categories good via non-financial
channels. Gopinath et al. (2012) studies the trade price fluctuation in different categories
of goods during 07 banking crisis. Engel and Wang (2011) provide an insight to the links
between business cycle, trade volatility and durable goods.
2.3 Model
2.3.1 Background context
Assume a world with N countries and M varieties of goods. All consumers have