Section 1: Introduction
We are attempting to create a model for the U.S import demand from Japan. This model that we
are created will include variables that will explain the changes in the demand. The questions we
will address will explain how exchange rates, inflation, wealth, income and consumer sentiment
are all important components to the model. The purpose of this paper is to explain the model we
have created from Japanese imports.
Qxd=f(Pxi,pyi,ii,wi,ri,csi,ε i)
Where the observation qx is the quantity of good x demanded. Px is the price of good x, Py
is the price of a related good y, I is real measure of consumer income of good x, M2 is a moving
average that is measure of wealth, R is a 3 year interest rate, CS is a measure of consumer
confidence, and E is a stochastic error term.
Our expected sign of Px, assuming ceteris paribus, is negative because the law demand states
an increase in Px give Py, Income, M2, Interest rates, and consumer sentiment, demand for good
y will increase. Our expected sign of Py is positive because as the price of Y increases assuming
all other variables stay the same, the demand for X will fall, assuming it is a normal good. Our
expected sign of I (income) is positive. This is because as Income increases, quantity demanded
of good X rises as well as described in the income effect. The expected sign of W (wealth) is
positive, this is because as wealth increases the ability to consume a variety of goods increases.
Our expected sign for R (interest rates) is positive because if the Fed raises interest rates foreign
investors increase the demand for the dollar looking for higher returns on there money. As the
demand for the dollar increases it gains purchasing power for foreign goods. There for an
increase in R creates and increase in quantity demanded of X. The expected sign for CS
(consumer sentiment) is positive because we assume that if expectation for the economy fall,
consumers cut back on spending and the quantity of demand for X will fall.
log yt=α 0+ log α 1pxt+ log α 2pyt+ log α 3it+ log α 4+ movav(w,3)t+ log α 5rt+ log α 6cst+ε t
Px represents a value for the U.S. dollar to Japanese Yen exchange rate. It represents the
purchasing power of 1 U.S. dollar of x Japanese yen. Its expected sign is negative. We chose
this value because it represents the price of Japanese imported goods compared to U.S. goods.
Py is a ratio of the US inflation rate to the Japanese inflation ration. We chose this value because
it represents the change in price of domestic goods vs. imported Japanese goods when inflation
either country changes. Its expected sign is positive. We chose real disposable personal income
as our measure of income because its represents the real purchasing power of individual due to
the fact that it is adjusted for inflation and less taxes. Its expected sign is positive
Explanatory variable W is equal to M2 that represents a proxy variable to show the measure of
personal wealth of the country. This will be shown as a 3month moving average and its
expected sign positive. We chose the Three Year Treasury Bill as our R because its represents an
realistic rate on a loan you would receive for a three month car loan which makes up a large
portion of imports from Japan. Its expected sign is positive. The variable of Consumes
Sentiment CS was chosen because it creates a variety of indexes that describes future
expectations of the economy. Its expected sign is positive. Taking the log linear regression gave
us a better regression equation that fit the data.
Regression 1
Dependent Variable: LOG(Y)
Method: Least Squares
Date: 12/04/08 Time: 11:46
Sample (adjusted): 1980M02 2008M09
Included observations: 215 after adjustments
Critique
Signs
• According to theory the sign for Px should be negative. Our regression matched what
theory dictated with 0.684304. The expected sign for Py is positive. The results were as
expected with 0.004446. The expected sign of I is positive which match our results with
1.370078. The expected sign of W is positive which did not match our results with 1.393883,
therefore we will run an auxiliary regression for this variable. The variable R has an expected
sign is positive which we also see in our results 0.153019. CS was expected to be positive as well
and this was correct also with 0.608338.
TStats
• All tstats were greater than the absolute value of 2 (statistically significant) except for
PY. This means that 5 of the explanatory variables contain data with more then 95% of the data
was with in the acceptance region thus making those data statistically significant. This means
we should reject the null hypothesis and accept the alternative hypothesis.
Magnitudes
•
Income and M2 are the only variable we found to be elastic, all others were inelastic. In the other
variables there are no significant outliers to offset the others.
PX PY I M2 R CS

log  0.0044 1.3700 1.3938 0.1530 0.60833
elas 0.6843 46 8 83 19 8
Coefficients of Variation
• The coefficient of variation is relevant because it is below .2 making the average error
less then two deviations from the mean.
801.751 8539.43
CV 5 7 0.093888
Adjusted Rsquared
• The adjusted R squared was 0.940103.. The Adjusted R Squared tell us the percentage of
variation of the depended variable The closer the r squared is to 1, the better the estimated
regression fits the actual data. This means that almost 95% of the regression function is fitted to
the actual data.
FStat
• The FStat of this regress was 560.796158 and smaller then the linear regression fStat.
This was the only exception we found that made the linear regression equation preferable.
The model is a time series model using monthly data with a frequency from 1980.1 to 2008.10.
The variables in the model are:
CS I PX PY W3 Y R
88.397687 6062.48208 145.231356 0.00160869 3976.88100 8471.81304 6.79187861
Mean 9 1 6 6 8 3 3
12.141322 1480.76192 49.1004090 0.00724572 1686.41116 2961.96228 3.21429969
Std. Dev. 2 8 6 3 1 7 8
Observations 346 346 346 345 344 345 346
Section 5: Results
log yt=α 0+ log α 1pxt+ log α 2pyt+ log α 3it+ log α 4+ movav(w,3)t+ log α 5rt+ log α 6cst+ε t
Substituted Coefficients:
The results of our log linear estimated regression function showed stronger results than
our linear regression. When comparing the two regressions we took the following into account:
Expected sign, magnitude, and statistical significance of each explanatory variable. Also, we
considered the Fstats and Adjusted RSquared to judge the strength of each overall equation.
Our results showed that the Log linear equation gave us all the correct expected signs according
to theory, whereas the linear gave us a wrong sign for Py. We compared the Tstats of each
variable. The log linear equation had more significant Tstatistics with an absolute value greater
than 2. Because the values were greater than 2, they proved they were statistically significant.
The magnitudes also showed in the log linear equation that each variable elasticities were closer
together meaning that they effect our dependent variable Y similarly. The variables W and
showed to have elasticities higher than 1 meaning that they are elastic making then effect the
demand for imports greater then the other variables. The coefficient of variation was 0.093888
and is relevant because it is below .2 making the average error less then two deviations from the
mean. The fstat was the only exception of all of the test that showed that the linear regression
equation was better then the log linear. Sine all of the other tests were in favor of the log linear
and they fstats were still quite close, we still preferred the log linear over the linear regression.
The problems that we encountered were mainly focused on our Py. The log equation only
accounted for 215 observations out of the overall 346 observations, which is only 62% of the
data. This problem is a result of our Py being a ratio of US inflation rate to JP inflation rate. In
the months where the ratio of inflation was 0 in Japan, those months were excluded from the
results. We believe that this can be solved with a dummy variable. There were no other instances
where we would have had to address a problem. For example, wrong signs or non significant t
stat which would have required us to run an auxiliary regression and a Wald Test.
.4 10.4
.3 10.0
.2 9.6
.1 9.2
.0 8.8
.1 8.4
.2 8.0
.3 7.6
.4 7.2
80 84 86 88 90 92 96 98 00 02 04 06 08
Section 6 :Conclusion
To conclude our findings we collected and analyzed this data to assess the importance of
the import demand from Japan to the United States. The United States imports a significant
portion of its durable goods including automobiles and parts, electronics, and textile goods
from Japan. A change in the demand for Japanese good in America would drastically alter
the way we consume. Similarly if the cost of consuming Japanese goods increased it would
significantly alter the supply and demand of domestically produced goods. According to our
model Income and Wealth are the most significant determinates of the amount of goods we
import from Japan, along with a relative importance placed on the exchange rate and
consumer sentiment. A policy implication that will affect this would be an aggressive
strengthening of the dollar by increasing interest rates. This would: a) increase the M2 by
creating a larger number of demand deposits as seen in the W variable; b) create a greater
demand for U.S. investments, there for creating a greater demand for the dollar and driving
up exchange rates, as seen in the Px varriable. This increase in the exchange rate causes
purchasing power of the U.S.D. in Japan to rise and c) it would stem inflation in the U.S.
there for again increasing the purchasing power of the U.S. as can bee seen in the py variable.