0% found this document useful (0 votes)
49 views

Batch 2 - Group 9

The document analyzes the impact of various macroeconomic variables on per capita GDP in the United States from 1950 to 2000. It finds that consumer price index, real consumption expenditures, real investment by private sector, real government expenditures, and unemployment rate significantly affect per capita GDP, while real disposable personal income has a limited influence. The regression results show that consumer price index, real consumption, real disposable income, real government expenditures, and real private investment positively impact per capita GDP, while unemployment rate has a negative impact.

Uploaded by

Javed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
49 views

Batch 2 - Group 9

The document analyzes the impact of various macroeconomic variables on per capita GDP in the United States from 1950 to 2000. It finds that consumer price index, real consumption expenditures, real investment by private sector, real government expenditures, and unemployment rate significantly affect per capita GDP, while real disposable personal income has a limited influence. The regression results show that consumer price index, real consumption, real disposable income, real government expenditures, and real private investment positively impact per capita GDP, while unemployment rate has a negative impact.

Uploaded by

Javed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

9 IMPACT OF MACROECONOMIC VARIABLES ON

PER CAPITA GDP OF USA


Econometric methods of decision making/ Batch 2 / 26 Nov 2017

GROUP 9 ABSTRACT

Dhruv Consul (I010) Gross Domestic Product (GDP) being an


Digvijay Khichi (I024) important indicator of economic activity, is often
Shubhangani Maheshwari used by decision makers to plan economic policy.
(I033) The scope of this paper is to provide analysis of
Varun Talikoti (I053) macroeconomic factors that impact the per capita
Ramansh Tyagi (I057) GDP of USA. We have analyzed the long-term
relationship between per capita GDP vis-à-vis
change in consumer price index, real consumer
expenditures, real investment by private sector,
real government expenditures, real disposable
personal income and unemployment rate.
This study uses quarterly data from Department
of commerce, U.S Bureau of Economic Analysis
(BEA) from January 1950 to December 2000.
The empirical results reveal that consumer price
index, real consumption expenditures, real
investment by private sector, real government
expenditures, unemployment rate affects the per
capita GDP, while on the hand, the influence of
real disposable personal income is up to limited
extent only.

INTRODUCTION

Gross Domestic Product (GDP) of a country is the money value of all final goods and services
produced by all the enterprises within the borders of a country in a year. It represents the
aggregate statistic of all economic activity. The performance of economy can be measured with
the help of GDP. Forecasting future economic outcomes is an important component of the
decision-making process in central banks for all countries. Monetary policy decisions affect
the economy with a delay, so, monetary policy authorities must be forward looking, i.e. must
know what is likely to happen in the future. Gross domestic product (GDP) is amongst one of
the most important indicators of national economic activities for countries. Scientific prediction
of the indicator has important theoretical and practical significance on the development of
economic development goals.

Understanding macro dynamics of Indian economy may be useful for policy makers, traders
and investors. Results may reveal whether the change in GDP is the outcome of movement in
macroeconomic dimensions of the economy or something else. We analyzed the long-term
relationship between GDP and certain macroeconomic variables using the regression equation
model in order to establish relation between these factors. The empirical results reveal that
consumer price index, real consumption expenditures, real investment by private sector, real
government expenditures, unemployment rate affects the per capita GDP, while on the hand,
the influence of real disposable personal income is up to limited extent only. Figure 1 shows
GDP per person for the United States between 1870 and the present.

Figure 1: Per capita income over the years of United states of America

Source: U.S. Bureau of Economic Analysis

OBJECTIVE OF THE STUDY

The paper aims at the following objectives:

1) To explore the major macro-economic variables


2) To analyze the correlation among the macro-economic variables
3) To study the impact of chosen macro-economic variables on the per capita GDP of USA
REVIEW OF LITERATURE

Our hypothesis was intended for a long-term examination of the relationship between economic
growth and macro-economic variables as our regression models measured GDP per capita
growth over a course of five decades from 1950 to 2000. Research severely lacks examination
of the long-run relationship between per capita GDP and the macro-economic variables,
therefore. there is little guidance beyond the short and medium-run effects. The stability of the
growth rate is remarkable and surprising, with GDP per person lying close to a linear time trend
with a slope of just under 2 percent per year. GDP per person for the United States between
1870 and the present is shown in Figure 1. Despite the impressive fit of a linear trend, growth
has at times deviated noticeably from a 2-percent baseline. It is clear that growth was slower
pre-1929 than post-1950. Growth from 1950 to 1973 was comparatively faster (2.50), but then
slowed markedly until 1995 (1.82). The U.S. experience may also understate uncertainty about
the future, since other countries have often seen level as well as growth rate changes.

The technique for obtaining statistical estimates and inferences related to time series using
linear regression is applicable only to stationary series, as Granger and Newbold (1967) showed
forty years ago. Two or more nonstationary series can be regresses only in the case when there
exists a cointegrating relation between them (Hendry & Juselius, 2001).

There are many problems associated with the theory of long-term economic growth. The study
of Galor (2005) describes the evolution of income per capita since the epoch of Malthusian
stagnation and discusses the process, which induced the transition to the current sustained
economic growth in developed countries. The primary aim of Galor’s study, as well as
economics as a whole, is to find a unifying theory accommodating various periods of growth
based on solid micro foundations. This is a fundamental approach which should be also
supported by observations at macro level.

We do not use any sophisticated technique of signal extraction, as proposed by Pedregal (2005),
who explored two linear trend models with a nonlinear forecast function. Under our framework,
the long-term forecast is not limited in time but is based on a constant annual increment of real
GDP per capita, which we call the inertial growth due to its resemblence to inertia in classical
mechanics. In other words, a developed economy will grow at a constant annual increment of
real GDP per capita, when no external forces are applied. Here, the model does not permit an
exponential growth path, unlike that presumed in the trend extracting procedures developed by
Pollock (2007) or in the original Solow (1956) and Swan (1956) growth model.
Mathematically, the Solow–Swan model is a nonlinear system consisting of a single ordinary
differential equation which models the evolution of the per capita stock of capital. Solow–
Swan proved to be a convenient starting point for various extensions due to its particularly
attractive mathematical characteristics. For instance, David Cass and Tjalling Koopmans in
1965 integrated Frank Ramsey's analysis of consumer optimization, thereby endogenizing
the savings rate, to create what is now known as the Ramsey–Cass–Koopmans model.
RESEARCH METHODOLOGY:

The study is focussed on six major variables i.e. real consumption, real government
expenditure, real investment, unemployment rate, real disposable income and consumer price
index. We have studied the impact of these macro-economic variables on the per capita GDP
of US. For this analysis, we have taken the quarterly data of all these variables from 1950 to
2000. We then employed a multiple regression model to test the effects of the identified
macroeconomic factors on the GDP of the country. Only secondary data has been used for
our analysis.

With our independent variables, the prediction of Y is expressed by the following equation:

lnpercapgdp = b0 + b1 lnrealcons+ b2 lncpi + b3 lnrealgovt +b4 lnrealinvs +b5 lnRealDpi


+ b6 lnunemp

Lnpercapgdp = log(Per capita gdp )


Lnrealcons = log(real consumption)
Lncpi = log(consumer price index)
Lnrealgovt = log(real government expenditure)
Lnrealinvs = log(real investment)
LnRealDpi = log(real disposable income)
Lnunemp = log(Unemployment rate)

With this basic functional form, we have performed the regression analysis. Following are the
processes which have been done to arrive at the best regression model:

Stationarity:

Since, we have a time series data, the stationarity of the dependent and independent variables
is primarily checked using Augmented Dickey-Fuller test.

The research had the following findings:

The per capita real GDP, real consumption and real disposable income have a first difference
stationarity. The consumer price index, real government expenditure, real private investment,
unemployment rate have trend stationarity.

As all the independent variables are either first difference stationary or trend stationary, we
use the functional form as it is for further study.
After we ran the regression, we got the following output:

Table 1: First regression output with autocorrelation

Variable Coefficient Std. Error t-Statistic Prob.

LNCPI 0.055978 0.004152 13.48061 0.0000


LNREALCONS 0.136820 0.040785 3.354690 0.0010
LNREALDPI 0.338774 0.037960 8.924510 0.0000
LNREALGOVT 0.023584 0.011504 2.049994 0.0417
LNREALINVS 0.058893 0.010668 5.520532 0.0000
LNUNEMP -0.070751 0.004325 -16.35933 0.0000
C -1.527811 0.031516 -48.47716 0.0000

R-squared 0.999235 Mean dependent var 2.960822


Adjusted R-squared 0.999211 S.D. dependent var 0.317521
S.E. of regression 0.008916 Akaike info criterion -6.568175
Sum squared resid 0.015661 Schwarz criterion -6.454318
Log likelihood 676.9538 Hannan-Quinn criter. -6.522117
F-statistic 42874.11 Durbin-Watson stat 0.320111
Prob(F-statistic) 0.000000

We clearly see that Prob(F-statistic) is 0.00 which means that the independent variables taken
together are significant in explaining the model. But, we also see that the Durbin-Watson
statistic is 0.3201 which is a very low value indicating the presence of autocorrelation in our
model.

Autocorrelation:

As we detected an autocorrelation in our model, we did an autocorrelation removal remedy. To


check for autocorrelation we did the Breush-Godfrey serial correlation LM test as a residual
diagnostic and found a lag to be significant at residual(-1) as shown in the table.

Table 2: Breush-Godfrey serial correlation LM test output

Variable Coefficient Std. Error t-Statistic Prob.

LNCPI -0.000356 0.002239 -0.158952 0.8739


LNREALCONS 0.018709 0.022235 0.841429 0.4011
LNREALDPI -0.034080 0.020972 -1.625032 0.1058
LNREALGOVT 0.013753 0.006359 2.162717 0.0318
LNREALINVS 0.004698 0.005774 0.813712 0.4168
LNUNEMP 0.005735 0.002419 2.370290 0.0188
C -0.006919 0.017042 -0.405975 0.6852
RESID(-1) 0.801452 0.072071 11.12038 0.0000
RESID(-2) 0.048296 0.092435 0.522485 0.6019
RESID(-3) 0.030572 0.073574 0.415524 0.6782

We auto-regressed the model using the residual at lag 1 i.e. ar(1) which was found to be
significant. Table 3 shows the Durbin-Watson statistics post autocorrelation treatment comes
out to be 1.7169 which sufficiently signifies the removal of auto correlation from our model.
Table 3: Output without autocorrelation

Variable Coefficient Std. Error t-Statistic Prob.

LNCPI -0.048427 0.021582 -2.243903 0.0260


LNREALCONS 0.421406 0.036231 11.63124 0.0000
LNREALDPI -0.002264 0.028217 -0.080232 0.9361
LNREALGOVT 0.165289 0.012427 13.30032 0.0000
LNREALINVS 0.131793 0.005488 24.01349 0.0000
LNUNEMP -0.014479 0.004269 -3.391536 0.0008
C -2.020023 0.177514 -11.37952 0.0000
AR(1) 0.995226 0.012158 81.85558 0.0000
SIGMASQ 9.08E-06 6.98E-07 13.00850 0.0000

R-squared 0.999910 Mean dependent var 2.960822


Adjusted R-squared 0.999906 S.D. dependent var 0.317521
S.E. of regression 0.003082 Akaike info criterion -8.660673
Sum squared resid 0.001852 Schwarz criterion -8.514285
Log likelihood 892.3886 Hannan-Quinn criter. -8.601457
F-statistic 269340.4 Durbin-Watson stat 1.716957
Prob(F-statistic) 0.000000

Post removal of autocorrelation, we saw that the real disposable income turned out to be highly
insignificant. Therefore, we dropped this variable and ran the regression to establish that all the
other variables are significant for the model. Output can be seen in Table 4.

Table 4: Output with significant independent variables

Variable Coefficient Std. Error t-Statistic Prob.

LNCPI -0.048619 0.021587 -2.252192 0.0254


LNREALCONS 0.420156 0.028292 14.85083 0.0000
LNREALGOVT 0.165062 0.012416 13.29466 0.0000
LNUNEMP -0.014524 0.004204 -3.454807 0.0007
LNREALINVS 0.131627 0.005133 25.64400 0.0000
C -2.024505 0.175310 -11.54811 0.0000
AR(1) 0.995301 0.012125 82.08748 0.0000
SIGMASQ 9.08E-06 6.85E-07 13.25257 0.0000

R-squared 0.999910 Mean dependent var 2.960822


Adjusted R-squared 0.999906 S.D. dependent var 0.317521
S.E. of regression 0.003074 Akaike info criterion -8.670445
Sum squared resid 0.001852 Schwarz criterion -8.540322
Log likelihood 892.3853 Hannan-Quinn criter. -8.617808
F-statistic 309410.3 Durbin-Watson stat 1.717306
Prob(F-statistic) 0.000000

Multicollinearity:

We further checked for the presence of multicollinearity by analysing the Variance Inflation
factor with the help of coefficient diagnostics test. We found that the absence of high
multicollinearity and the VIF values were in the lower range as shown in Table 5.
Table 5: Multicollinearity test output
Coefficient Uncentered Centered
Variable Variance VIF VIF

LNCPI 0.000466 30.59319 3.417177


LNREALCONS 0.000800 110.1053 4.394229
LNREALGOVT 0.000154 16.77015 1.819398
LNUNEMP 1.77E-05 3.450940 3.206549
LNREALINVS 2.63E-05 4.073432 2.366519
C 0.030734 70.60554 NA
AR(1) 0.000147 2.364553 1.940394
SIGMASQ 4.69E-13 1.248181 1.244805

Heteroscedasticity Test:

We also ran the Breusch-Pagan test to check for Heteroscedasticity. We found that there was
no heteroscedasticity with P value being 0.3993 as shown in Table 6.

Table 6: Heteroscedasticity test output

Heteroskedasticity Test: Breusch-Pagan-Godfrey

F-statistic 1.033098 Prob. F(5,198) 0.3993


Obs*R-squared 5.186706 Prob. Chi-Square(5) 0.3935
Scaled explained SS 10.29014 Prob. Chi-Square(5) 0.0674

CONCLUSION:

1. Per capita real GDP is dependent on real consumption, real private investment, real
government expenditure, real disposable personal income and unemployment rate.

2. Real disposable personal income was found to be insignificant and hence was dropped from
the model.

3. Adjusted Square is 0.99 implying 99% of variation in dependent variable is explained by


independent variables.

4. Regression Equation:

Log(percapgdp)= -2.204505 + 0.4202*log(RealCons) + 0.131627*log(RealInvs)


+0.1650*log(RealGovtExp) - 0.048619*log(cpi) -0.0145*log(unemp) + 0.995*ar(1)

5. A Positive consumption coefficient indicates that an increase in consumption will increase


the GDP which is true as the consumption increases, industries will employ a larger workforce
and in turn increasing incomes.

6. An investment in the economy will create more opportunities for economic growth.

7. The coefficient of government expenditure is positive which indicates that any expenditure
made in health care, education and infrastructure will have a positive impact on GDP.
8. The negative coefficient of unemployment indicates that the rise in employment levels will
results in an increase in GDP of the country.

REFERENCES:

1) Clive Granger and P. Newbold (1967) ‘Spurious regressions in econometrics’, Journal of


Econometrics, vol. 2, issue 2, 111-120.
2) D. F. Hendry and K. Juselius (2001) ‘Explaining Cointegration Analysis Part II,’ Energy
Journal, Vol. 22, No. 1, 2001, pp. 75-120.
3) Oded Galor (2005) ‘From Stagnation to Growth: Unified Growth Theory’, Handbook of
Economic Growth, Elsevier.
4) Angus Maddison (2004) ‘Quantifying and interpreting world development:
macromeasurement before and after Colin Clark’, Australian Economic History Review,
Economic History Society of Australia and New Zealand, vol. 44(1), pages 1-34, March.
5) Pedregal, D.J., Young, P.C. (2002), ‘Statistical Approaches to Modelling and Forecasting
Time Series’, Companion to Economic Forecasting (Blackwell Publishers), 69-104.
6) Solow, R.M. (1956) ‘A Contribution to the Theory of Economic Growth’, Quarterly
Journal of Economics 70 (1), 65 − 94.
7) Swan, T.W. (1956) ‘Economic Growth and Capital Accumulation’, Economic Record 32
(63), 334 − 361.

You might also like