Sie sind auf Seite 1von 27

1/30/2014

UOM
IS THE PHILLIPS CURVE ALIVE IN THE U.S?
Neoclassical Views


Submitted By:
Phoolchund Reshmee
Dorasamy Sebastien







ECON 2003Y (3) Principles of Econometrics
1

Table of Contents
1.0 Introduction .2
1.1 Statement of theory..3
2.0 Specification of the econometric model...4
2.1 Difficulties of estimating Phillips curve...........6
2.2 Assumption of our model..6
3.0 Obtaining the data for the United Sates and Empirical evidence.7
4.0 Estimation of the econometric parameters and hypothesis testing..10
4.1 Uncorrelated Phillips curve in U.S 1970s....11
4.2 Phillips curve chow test...12
4.3 Aftermath of 1970s..13
4.4 Recession 2008 and its impact on Phillips curve.13
4.5 Dummy variables to detect structural change: Recession14
4.6 Linear model of Phillips curve (2000-2013)15
5.0 Multiple Regression.16
5.1 When to add a new variable.16
5.1.1 The marginal contribution of Imports...17
5.1.2 The marginal contribution of Real GDP Growth Rate.18
5.1.3 The marginal contribution of real interest rates....19
5.2 Testing the overall significance f the model....20
6.0 Relaxing the assumption.....21
6.1 Multicollinearity..............................21
6.2 Autocorrelation22
6.2.1 Detecting autocorrelation..22
6.3 Heteroscedasticity........23
6.3.1 Detecting heteroscedasticity.....24
7.0 Forecasting...24
8.0 Using the model for control or policy purposes..25
9.0 Conclusion...26
10.0 References..26
ECON 2003Y (3) Principles of Econometrics
2

1.0 INTRODUCTION
Inflation and unemployment are two key features of an economy. Inflation is simply a general
increase in the average prices of goods over time and thus, leading to a fall in the purchasing
power. Main causes of inflation are demand pull and cost push. It can be measured using
Consumer Price Index (CPI). On the other hand, unemployment is the state of being unemployed
when a worker is actually searching for a job. Causes could be structural, cyclical or even
seasonal. One method of measuring unemployment rate is by using claimant count, that is, the
number of people receiving unemployment benefits. High values of both show the poor health
of an economy because they are the main key economic indicators.
The opportunity cost of lower unemployment is higher inflation and vice versa. Thus, it is said
that there exists a tradeoff between these two variables. This particular relationship between
inflation and unemployment is known as the Phillips Curve. Our purpose in doing this report is
to provide empirical evidence in the United States by having recourse to regression analysis. We
have chosen the latter with a neoclassical school of thought with backward linear expectations. It
is one having rational expectations. Throughout we will test if the present day Phillips Curve is
consistent with the neoclassical beliefs.
Part one is about theory and the other sections will concentrate on earlier and recent empirical
evidence. What kind of relationship prevails? Inverse, positive or no relationship at all? Is the
inflation rate or the unemployment the dependent variable? Well, it all depends on the researcher
and weve decided to make the inflation rate the dependent variable. [The lower the
unemployment rate the more people will demand more goods via an increase in wages leading to
inflation]. Can we forecast inflation rate given the value of known unemployment rate? What
policy must be used to exploit the trade off?
To answer all these we will briefly go into the history of the Phillips Curve and know the
different variables in the equation of the curve to be able to make statistical inferences and
answer all these issues. However, there can be limitations but we will try as much as possible to
resolve the problems such as multicolliearity, heteroscedascity and autocorrelation.

ECON 2003Y (3) Principles of Econometrics
3

1.1.STATEMENT OF THEORY
In 1958, A.W. Phillips a New Zealand born economist, wrote a paper titled The Relation between
Unemployment and the Rate of Change of Money Wage Rates in the UK, 1861-1957, which was
published in the quarterly journal Economica and showed that nominal wage growth was
negatively correlated with unemployment in the United Kingdom.
Hence, the Phillips curve shows the short-run trade-off between inflation and unemployment.
= -f(U)
Two years later, in 1960, Paul Samuelson and Robert Solow found a negative correlation
between United Sates inflation and unemployment rates, naming it the Phillips Curve.
It is deduced that there are two frames for the Phillips curve:
- The short-run Phillips curve : an increase in inflation would lower unemployment
- The long-run Phillips curve : no trade off- positive relationship
These are shown in figure 1 and 2 below:

Figure 1 Figure 2
ECON 2003Y (3) Principles of Econometrics
4

Figure 1 shows the short run Phillips Curve (SRPC) which is drawn on the assumption of a given
expected rate of inflation. So if there were an increase in inflation caused by a monetary
expansion and this has the effect of driving inflationary expectations higher causing an upward
shift in the short run Phillips Curve. At low aggregate demand, inflation is low and since labour
is a derived demand, demand for labour is low, therefore unemployment is high.
Figure 2 shows the natural rate of unemployment hypothesis which states that unemployment
eventually returns to its natural rate, regardless of the inflation rate based on the classical
dichotomy (long run aggregate supply-LRAS). An inward shift in the long run Phillips Curve
might be due to supply-side improvements to the economy and a reduction in the natural rate of
unemployment.

2.0 SPECIFICATION OF THE ECONOMETRIC MODEL
To bridge the gap between theory and evidence, Friedman and Phelps introduced a new variable:
expected inflation a measure of how much people expect the price level to change. The
econometric model we will be using is:
t
n
t t
e
t t
u U U + [ = [ ) ( |
, where





Given expected inflation, t
t
e
, an increase in the unemployment rate leads to a decrease in
inflation, t
t.
The equation above denotes the inverse relationship between inflation rate and
unemployment rate. The natural rate of unemployment is sometimes referred to as the Non
Accelerating Rate of Inflation (NAIRU) where inflation rate remains stable.

Actual
inflation rate
at time t

Inflation depends on expected inflation :
being formed in year (t 1)

Error Term: e.g.
supply shocks

The deviation of
unemployment
from the natural
rate

ECON 2003Y (3) Principles of Econometrics
5

The other variables in the equation are notably:
U: Unemployment Rate
U
n
: Natural Rate of Unemployment
U-U
n
: Unemployment Gap
: measures the response of inflation to cyclical unemployment: the rate of change in the average
value of t
t
.
It is to be noted that this equation is derived from the Aggregate Supply equation.

Writing the Neoclassical Phillips curve equation:

t
=
e
t
- (U
t
U
t
n
) + u
t
.(1)

t

e
t
= - (U
t
U
t
n
) + u
t
.(2)

e
t
=
t1
: the inflation expected this year is the inflation rate that prevailed in the last year.
This is known as the random walk.

Rewriting the regression model in standard form, we obtain the following equation:

t

t1
= -U
t
+ U
t
n
+ u
t..(3)

t
= -U
t
+

+ u
t..(4)

t
= - U
t
+u
t..(5)

Last equation states that the change in the inflation rate between two time periods is linearly
related to the current unemployment rate. Throughout this report we will be using equation 5.
ECON 2003Y (3) Principles of Econometrics
6

2.1. DIFFICULTIES OF ESTIMATING PHILLIPS CURVE
- Estimation of
e
t
and U
t
n

Such variables are difficult to estimate in real life.
- Dummy Variables: since the 1970s it has become common practice to include dummy
variables and measures of supply shocks, such as the relative price of energy. Shocks to
such variables arguably create positive correlation between ination and unemployment,
and would thus, bias the estimate of if omitted.
- Accuracy of Data: Maybe the quality of data collected over the internet may be poor.
The success of regression analysis depends on the availability of the appropriate data.

2.2. ASSUMPTION OF OUR MODEL
- Wages and prices are completely flexible with respect to changes in expected inflation.
- E(u
t
)=0 The errors have zero mean
- Var(u
t
)=
2
- The variance is constant and finite over all values of X
t

- Cov (u
i
,u
j
)=0 - Errors are statiscally independent of each other: No autocorrelation
- Cov(u
t
, X
t
) =0 Zero covariance
- Normally Distributed-To make statistical inferences about the population parameters
from the sample data. It will follow a t-distribution.
- The regression model is correctly specified- No specification bias or error
- There is no perfect multicollinearity- No perfect linear relationships among the
explanatory variables.
- The regression would be linear in parameters.

ECON 2003Y (3) Principles of Econometrics
7

3.0 OBTAINING THE DATA FOR THE UNITED STATES AND EMPIRICAL EVIDENCE
We have collected data for the United States on inflation rate and unemployment rate on an
annual basis from 1960 to 2013. The size of the data is 54. The source is from Bureau of Labor
Statistics. Inflation rate is measured by CPI. The unemployment rate is the civilian
unemployment. Both are collected using monthly censuses.
The summary results are as follows:
Unemployment Rate: X Inflation Rate: Y
n=54 n=54
Mean: 6.10 Mean=3.96
Mode: 5.5 Mode=1.6
Median: 5.75 Median=3.2
Std Deviation=1.59 Std Deviation=2.81
Kurtosis: -0.17 Kurtosis= 2.62


On average unemployment rate was 6.1% and inflation rate was 3.96 over the last 54 years. The
point estimators namely mean, mode and median is higher for unemployment rate. Standard
deviation which is the measure of variability is higher for inflation rate by 1.22 thus, slightly
further away from the mean. Kurtosis is a measure of whether the data are peaked or flat relative
to a normal distribution. The kurtosis of unemployment is less than zero indicates a relatively flat
distribution compared to inflation rate.
Note: Test of normality: Kurtosis

Normally distributed value: 3
However, limitations of time series data is that it assumes data to be stationary, that is, mean and
variance do not vary systematically over time.
5 . 329 = x 214 = y
2 2
4
] ) ( [
) (

=
x E
x E
k
ECON 2003Y (3) Principles of Econometrics
8

Plotting inflation rate and unemployment rate against time we get the graph below:

Figure 3
There has been some inverse relationship as well as positive ones over the last 54 years.
Between 1979 and 1983, we see inflation fall from 15% to 2.5% and a rise in unemployment
from 5% to 11%.This suggests there can be a tradeoff between unemployment and inflation.
In 2008, the recession caused a sharp rise in unemployment and inflation became close to zero.
Plotting the same data set but inflation rate against unemployment rate we get the resulting
Phillips Curve.
Note: We will use the level form because the percentages are small values. Therefore, we would
not be using log in our approach. Only the linear neoclassical functional form will be tested
throughout.
The Phillips Curve for the period 1960-2013 referred as population regression function is shown
on the next page.
-2
0
2
4
6
8
10
12
14
16
1
9
6
0
1
9
6
2
1
9
6
4
1
9
6
6
1
9
6
8
1
9
7
0
1
9
7
2
1
9
7
4
1
9
7
6
1
9
7
8
1
9
8
0
1
9
8
2
1
9
8
4
1
9
8
6
1
9
8
8
1
9
9
0
1
9
9
2
1
9
9
4
1
9
9
6
1
9
9
8
2
0
0
0
2
0
0
2
2
0
0
4
2
0
0
6
2
0
0
8
2
0
1
0
2
0
1
2
U.S Inflation and Unemployment Rates (1960-2013)
Inflation Rate (%) Unemployment Rate (%)
ECON 2003Y (3) Principles of Econometrics
9



Figure 4 Figure 5
It is important to note that such a scatter graph may enable a labour economist to predict the
average change in inflation rate given a certain unemployment rate.
Figure 4 which is from 1960-2013, we do not see the well-known inverse relationship.
But if we break down the same graph into shorter time periods we are able to see downward
sloping graphs which are shown in figure 5.
The US data from 19601969 and 1885-1990 both suggest a trade-off between inflation and
unemployment rates.
However, this is not always the case and it will be shown later that there can also be positive
relationship between inflation and unemployment rates.



y = 0.1654x + 2.9538
R = 0.0088
-2
0
2
4
6
8
10
12
14
16
0 5 10 15
I
n
f
l
a
t
i
o
n

R
a
t
e

(
%
)

Unemployment Rate (%)
Population Regression Function
y = -1.181x + 8.0014
R = 0.7133
y = -1.1562x + 10.995
R = 0.605
0
1
2
3
4
5
6
7
0 5 10
I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate (%)
Sample Regression Functions
1960-1969
1985-1990
No trade off in the LR SR Trade off

ECON 2003Y (3) Principles of Econometrics
10

4.0 ESTIMATION OF ECONOMETRIC PARAMETERS AND HYPOTHESIS TESTING
We start from earlier days when inflation rate was negatively related to unemployment rate. The
sample consists of data from 1960-1969. As depicted, this is a downward sloping Phillips curve
represented by the linear functional form.

Figure 6
Linear model:
t
=8.0014 1.1802U
t
+ u
t
df=8
SE = (1.29) (0.265) F(1,8)=19.90
t = (6.18) (-4.46) r
2
= 0.7133
p = (0.000) (0.002)
Regression results show that if the unemployment rate goes down by 1 percentage point, on
average, the inflation rate goes up by about 1.18%, and vice versa. The intercept term shows that
if unemployment rate is zero then the mean value of inflation rate is about 8%.
The p values are significant at 5% significance level implying there is enough evidence against
Ho.
The F value is also significant at 5% significance level.
r which measures the percentage of the total variation of inflation rate is about 71% is
explained by unemployment rate. To sum up in the 1960s the sample supported the views of
Williams Phillips.
However, it is interesting to note that it is not always the case when the Phillips curve slopes
downwards. There have been many occasions where this belief was broken.
y = -1.1802x + 8.0014
R = 0.7133
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8
I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate (%)
ECON 2003Y (3) Principles of Econometrics
11

4.1. UNCORRELATED PHILLIPS CURVE IN U.S IN 1970S

Figure 7
In the 1970s, however, the relation broke down. In the USA and most other OECD countries,
there was both high inflation and high unemployment, clearly contradicting the original Phillips
curve. Hence, the relation broke down. This is a situation of stagflation where both the variables
are increasing.
The reasons were simply because of supply shocks. The United States was hit twice in the 1970s
by large increases in the price of oil & rising expected inflation worsened the tradeoff. This
could be referred to as structural change that is, the values of the parameters do not remain the
same throughout the entire time period. Structural change can be a consequence of oil shocks,
food crisis or even financial crisis.
We will use the Chow Test to detect whether a structural change has occurred causing the
negative relationship to turn out into a positive one.





y = -1.1802x + 8.0014
R = 0.7133
y = 0.3291x + 5.0365
R = 0.0214
0
2
4
6
8
10
12
0 2 4 6 8 10
I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate(%)
1960-1969
1970-1979
ECON 2003Y (3) Principles of Econometrics
12

4.2. PHILLIPS CURVE CHOW TEST
Time period 1960-1979:
t
= 0.745-0.61U
t
+ u
t

n = 21 RSS=177.995


Time period 19601969:
1t
= 8.0014 1.1802U
1t
+ u
1t

n
1
= 10 RSS
1
=6.034
Time period 19701979
2t
= 5.037+0.333U
2t
+ u
2t
n
2
= 10 RSS
2
=63.04
The hypotheses would be as follows:
H
0
:
1t
=
2t
= 0
H
1
: at least one of
1t
or
2t
is non zero
If
1t
=
2t
= 0, then there is no difference between the two time periods. It assumes that the
intercept as well as the slope coefficient remains the same over the entire period.
If at least one of
1t
, or
2t
are nonzero, the regression function changes at date t.
The resulting F test would be as follows:

We reject Ho if Estimated F-value> Critical F-value at 5 % significance level.
RSS
UR
= RSS
1
+ RSS
2
with df = (n1 + n2 2k)
=6.034+63.04 =69.074
RSS
R
= 177.995
Estimated F=12.61
Critical F value for 2 and 17 df the 5% sig level is 3.59. Since 12.61> 3.59, we reject at 5 %
significance level implying that a structural shift has occurred. However a limitation is that the
null hypothesis allows for no differences at all between the groups.
)] 2 ( , [(
2 1
2 1
~
2 / ) (
/ ) (
k n n k
UR
UR R
F
k n n RSS
k RSS RSS
F
+
+

=
ECON 2003Y (3) Principles of Econometrics
13

4.3 AFTERMATH OF 1970

Figure 8 Figure 9
Afterwards, the data no longer show any sign of a negative relationship between inflation and
unemployment. A new relationship appeared in the form of change in inflation.
4.4 RECESSION 2008 AND ITS IMPACT ON PHILLIPS CURVE


Figure 10
y = 0.4181x + 0.6062
R = 0.1744
y = 0.1662x + 4.3415
R = 0.0047
0
2
4
6
8
10
12
14
16
0 5 10 15
I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate(%)
1980s and 1990s
y = -1.0984x + 7.3352
R = 0.5311
y = -0.301x + 1.4708
R = 0.2288
-5
-4
-3
-2
-1
0
1
2
3
0 5 10 15
C
h
a
n
g
e

i
n

I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate(%)
1980s and 1990s Using Change in
Inflation Rate
1980-1989
1990-1999
y = -0.2402x + 1.2852
R = 0.042
y = 0.2915x - 2.9845
R = 0.0112
-5
-4
-3
-2
-1
0
1
2
3
0 2 4 6 8 10 12
C
h
a
n
g

i
n

I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate(%)
Before Crisis
After Crisis
ECON 2003Y (3) Principles of Econometrics
14


_cons 1.327741 2.356822 0.56 0.584 -3.859589 6.51507
d1 .4906489 1.683161 0.29 0.776 -3.213964 4.195262
unemployme~e -.2486831 .4531738 -0.55 0.594 -1.246112 .7487458

dinflation Coef. Std. Err. t P>|t| [95% Conf. Interval]

Total 31.3149988 13 2.40884606 Root MSE = 1.6568
Adj R-squared = -0.1396
Residual 30.1966681 11 2.74515164 R-squared = 0.0357
Model 1.11833072 2 .559165361 Prob > F = 0.8187
F( 2, 11) = 0.20
Source SS df MS Number of obs = 14
reg dinflation unemploymentrate d1
. - preserve
. gen d1=(tin(2008,2013))
delta: 1 unit
time variable: year, 2000 to 2013
The only time it shown evidence of holding is before 2008 when unemployment increased
significantly and inflation decreased. However, from 2009 onwards the relationship did not held
well. In 2011 both inflation and unemployment fell and show little correlation and there are a
number of data points that could defy the theory. The Feds excessive monetary policy and the
fall in participation rate together are strong factors that can explain why the Phillips Curve
relationship has not held as of late. This will allow us to investigate whether the recent nancial
crisis can be considered a structural change in the Phillips curve.

4.5 DUMMY VARIABLES TO DETECT STRUCTURAL CHANGE: RECESSION
Time period 2000-2013:
t
=1.29+ 1.84D
t
-0.24U
t
-0.01(D
t
U
t
) + u
t

Insignificant t values and insignificant F test!!
R
2
= 0.0357
D=1 for observation 2008-2013 and 0 otherwise
The mean function for the period 2000-2007 is:
E(
t
/D
t
=0, U
t
)= 1.29 -0.24U
t

The mean function for the period 2008-2013 is:
E(
t
/D
t
=1, U
t
)= 1.84-0.25U
t

Using F test the null hypothesis is not rejected, the regression lines will coincident. Hence,
structural change has not happened in the form of recession. The results are below:




ECON 2003Y (3) Principles of Econometrics
15

4.6. LINEAR MODEL OF PHILLIPS CURVE (2000-2013)

Figure 11
Linear model:
t

t1
= 0.824 - 0.137U
t
+ u
t
df=12
SE = (1.54) (0.232) F(1,12)=0.35
t = (0.54) (-0.59) r
2
= 0.0283
p= (0.602) (0.566)

Regression results show that if the unemployment rate goes up by 1 percentage point, on
average, the change in the inflation rate goes down by about 0.137 percentage points, and vice
versa.
The intercept term shows that if unemployment rate is zero then the mean value of inflation rate
is about 8%.
The natural rate of unemployment is about 6.01%
All t values, p values and F test are insignificant leading us to conclude that the regression itself
is insignificant and the independent variable is not actually explaining any change in the inflation
rate.
y = -0.1368x + 0.8235
R = 0.0283
-5
-4
-3
-2
-1
0
1
2
3
0 2 4 6 8 10 12
C
h
a
n
g
e

i
n

I
n
f
l
a
t
i
o
n

R
a
t
e
(
%
)

Unemployment Rate(%)
ECON 2003Y (3) Principles of Econometrics
16

5.0. MULTIPLE REGRESSION
Inflation is also affected by other variables such as:
Labour costs
Real GDP growth rate
Import prices
Costs of production
Real interest rates
We shall test how the inclusion of these variables namely real GDP growth rate, import prices
and real interest rate will affect our dependent variable, inflation.


5.1. WHEN TO ADD A NEW VARIABLE
Researchers choose the model that gives the highest adjusted R
2
.
The question that arises is when adjusted R
2
increases?
- It will increase if the t value of the coefficient of the newly added variable is larger
than 1 in absolute value.
- Adjusted R
2
will increase with the addition of an extra explanatory variable only if
the F ( =t2) value of that variable exceeds 1.




ECON 2003Y (3) Principles of Econometrics
17

Mean VIF 1.13

unemployme~e 1.13 0.884719
imports 1.13 0.884719

Variable VIF 1/VIF
. vif

_cons -4.712516 4.287276 -1.10 0.295 -14.14875 4.723716
imports .4123889 .299615 1.38 0.196 -.2470594 1.071837
unemployme~e -.2477765 .237481 -1.04 0.319 -.7704686 .2749156

dinflation Coef. Std. Err. t P>|t| [95% Conf. Interval]

Total 31.3149988 13 2.40884606 Root MSE = 1.5362
Adj R-squared = 0.0203
Residual 25.9591459 11 2.35992236 R-squared = 0.1710
Model 5.35585287 2 2.67792644 Prob > F = 0.3564
F( 2, 11) = 1.13
Source SS df MS Number of obs = 14
. reg dinflation unemploymentrate imports
5.1.1. The Marginal Contribution of Imports.
Expectations: An increase in import prices will lead to higher imported inflation which will
thereby lead to a rise in demand for local goods and hence reduce unemployment in US and vice
versa.
The results are as follows:
Time Period: 2000-2013








Adding imports has increased our adjusted R
2
by 0.073. The incremental contribution of the new
variable can be conducted through F test:
F Test = =1.90 >1

ANALYSIS OF RESULTS:
Firstly, a positive relationship was found between inflation and import (0.41) as expected.
Secondly the t value of the coefficient of the newly added variable is larger than 1. (1.38).
Moreover the adjusted R
2
has also increased by adding imports(From 0.0527 to 0.0203) This
may be due to the absolute t value and F test which is also greater than 1. Next the VIF is less
than 10 (1.13) and tolerance ratio is not that near 0 (0.88), so low multicollinearity exist, It would
therefore be wise to include this variable in our model.
d new
n old new
df RSS
df ESS ESS
/
/
ECON 2003Y (3) Principles of Econometrics
18

Mean VIF 1.01

unemployme~e 1.01 0.989995
realgdpgro~e 1.01 0.989995

Variable VIF 1/VIF
. vif

_cons .1932341 1.645667 0.12 0.909 -3.428855 3.815323
realgdpgro~e .2633395 .250915 1.05 0.316 -.2889207 .8155998
unemployme~e -.1124676 .231738 -0.49 0.637 -.6225194 .3975842

dinflation Coef. Std. Err. t P>|t| [95% Conf. Interval]

Total 31.3149988 13 2.40884606 Root MSE = 1.5857
Adj R-squared = -0.0439
Residual 27.6601832 11 2.51456211 R-squared = 0.1167
Model 3.65481557 2 1.82740779 Prob > F = 0.5053
F( 2, 11) = 0.73
Source SS df MS Number of obs = 14
. reg dinflation unemploymentrate realgdpgrowthrate
5.1.2. The Marginal Contribution of Real GDP Growth Rate:
Expectations: The higher the real GDP growth rate the higher the inflation.
The results are as follows:
Time Period: 2000-2013









Adding real GDP growth rate has increased adjusted R
2
by 0.0088. Incremental contribution of
explanatory variable is denoted by the F test.
F Test = 1.10 > 1

ANALYSIS OF RESULTS:
There exists a positive relationship between the two (0.26) as expected. Then the t value of the
coefficient of the newly added variable is larger than 1. (1.05). In addition, The F value of this
variable exceeds 1. Thus the adjusted R
2
will also increase.(From 0.0527. to 0.0439). Next
the VIF is less than 10 and tolerance ratio is not near 0 (0.98) , so low multicollinearity exist, It
would therefore be wise to include this variable in our model.
ECON 2003Y (3) Principles of Econometrics
19

Mean VIF 2.07

unemployme~e 2.07 0.483085
realintere~s 2.07 0.483085

Variable VIF 1/VIF
. vif

_cons 2.027872 3.169384 0.64 0.535 -4.947894 9.003638
realintere~s -.1689663 .3843996 -0.44 0.669 -1.015024 .6770915
unemployme~e -.2458058 .3449402 -0.71 0.491 -1.005014 .5134025

dinflation Coef. Std. Err. t P>|t| [95% Conf. Interval]

Total 31.3149988 13 2.40884606 Root MSE = 1.6488
Adj R-squared = -0.1286
Residual 29.904669 11 2.71860627 R-squared = 0.0450
Model 1.41032982 2 .70516491 Prob > F = 0.7761
F( 2, 11) = 0.26
Source SS df MS Number of obs = 14
. reg dinflation unemploymentrate realinterestrates
5.1.3. The Marginal Contribution of Real Interest Rates
Expectations: The higher the expected inflation the higher the real interest rates.
Time Period: 2000-2013






Adding real interest rates has decreased adjusted R
2
by - 0.0759: Incremental contribution of
explanatory variable
F- Test = =0.1932 < 1
Any analysis of the Phillips Curve without the inclusion of real interest rates may generate a
misleading result as any changes in the interest rate impacts the labour capital input mix in the
production process, leading to a change in the level of employment in the economy. It is
important to include interest rate in our model.
ANALYSIS OF RESULTS:
From the above results, it can be seen that due to significant multicollinearity the coefficient of
interest rate is negative when according to theory it should have been positive. The tolerance
ratio is near 0. The adjusted R
2
has reduced significantly (F test and absolute t value less than 1).
Here it would be preferred to drop the collinear variable as it would be impossible to isolate the
individual impacts.
The various variables have yield different results. However in our model it is highly favourable
to include only those which are not highly correlated such as import prices and real GDP growth
rate.
d new
n old new
df RSS
df ESS ESS
/
/
ECON 2003Y (3) Principles of Econometrics
20

5.2. TESTING THE OVERALL SIGNIFICANCE OF THE MODEL
In fact it is important to further test whether it is favorable to add both imports prices and real
GDP in our model including the first variable unemployment.
Under assumptions of:
Disturbances u
t
are normally distributed (CNLRM)
H
0
: all slope coefficients are simultaneously zero
Then the F ratio is given by: ESS/k-1
RSS/n-k
If F > F
(k 1, n k),
reject H
0
( all slopes are not equal to zero)

MULTIPLE REGRESSION RESULTS:
Regression Eq
n
:
t

t1
= +
1
U
1t
+
2
X
2t
+
3
X
3t
+ u
t
X2t =Real GDP Growth Rate
X3t =Imports (% of GDP)
Results after Regression:

Mean VIF 1.13

realgdpgro~e 1.06 0.942715
unemployme~e 1.13 0.884334
imports 1.19 0.842466

Variable VIF 1/VIF
. vif

_cons -6.835859 4.305209 -1.59 0.143 -16.42846 2.756744
imports .5075635 .2910868 1.74 0.112 -.1410183 1.156145
realgdpgro~e .3533293 .2361583 1.50 0.165 -.1728642 .8795229
unemployme~e -.2407494 .2251935 -1.07 0.310 -.7425119 .261013

dinflation Coef. Std. Err. t P>|t| [95% Conf. Interval]

Total 31.3149988 13 2.40884606 Root MSE = 1.4564
Adj R-squared = 0.1195
Residual 21.2110896 10 2.12110896 R-squared = 0.3227
Model 10.1039092 3 3.36796973 Prob > F = 0.2533
F( 3, 10) = 1.59
Source SS df MS Number of obs = 14
. reg dinflation unemploymentrate realgdpgrowthrate imports
ECON 2003Y (3) Principles of Econometrics
21

Mean VIF 1.13

realgdpgro~e 1.06 0.942715
unemployme~e 1.13 0.884334
imports 1.19 0.842466

Variable VIF 1/VIF
. vif
t t1 = -6.834 -0.241U1t + 0.353X2t +0.508X3t + ut
SE = (4.3) (0.23) (0.24) (0.29) F(3,10)=1.59
t = (-1.59) (-1.07) (1.5) (1.74) R
2
= 0.3227
p= (0.143) (0.31) (0.17) (0.11) df=10

F value= 1.59
Critical value: 3.71
Since 1.59 < 3.71 we do not reject the null hypothesis and conclude that all slope coefficients
are zero and conclude that there is no evidence that any of the predictors are linearly associated
to Y. This is why it would have been better to make use of the reciprocal method which is a
reason why maybe these results tend to be insignificant.
6.0 RELAXING THE ASSUMPTION
6.1 MULTICOLLINEARITY
As discussed earlier we had so many insignificant values. Does it imply the sample is not good
or have we missed some variables?
When two X variables are highly correlated, they both convey essentially the same information.
It misleadingly inflates the standard errors. Thus, it makes some variables statistically
insignificant while they should be otherwise significant. It is detected by using Variance Inflation
Factor (VIF) or Tolerance Ratio (1/VIF). Our sample has low degree of multicolinearity.




ECON 2003Y (3) Principles of Econometrics
22

Durbin-Watson d-statistic( 4, 14) = 2.568728
6.2. AUTOCORRELATION
Recall: Cov (u
i
,u
j
)=0
- It occurs in time series data when the errors associated with a given time period carry
over into future time periods. Cov (u
i
,u
j
) =/= 0
- Cannot use t and F test
Some Causes of Autocorrelation
- Excluded Variable
- Nonstationary mean and variance
- Lags
We used
t =

t

t1 :
One of the explanatory variables is the lagged value of the
dependent variable: Autoregression

6.2.1. DETECTING AUTOCORRELATION
A. Durbin Watson Test defined as:

Assumptions are:
1. No lagged version of dependent variable among explanatory variable.
2. There is an intercept term in the regression model.
The regression result for autocorrelation is as follows:


=
=
=
=

=
n t
t
t
n t
t
t t
u
u u
d
1
2
2
2
1

) (
ECON 2003Y (3) Principles of Econometrics
23

H0: no serial correlation

1 1.555 1 0.2124

lags(p) chi2 df Prob > chi2

Breusch-Godfrey LM test for autocorrelation
B. Breusch-Godfrey lm test allows for:
1. Non schotastic regressors such as the lagged values of the regressand.
The result using this approach is:





However, using Durbin Watson Test to detect autocorrelation revealed a value near 2 forcing us
to conclude there is no serial correlation. A similar result is obtained using Breush Godfrey
Langrange Multiplier Test.



6.3. HETEROSCEDASTICITY
Recall: Var(u
t
)= - The variance is constant and finite over all values of X
t

- It is a case when the variance is not constant
- Heteroscedasticity causes standard errors to be biased.
- OLS estimators are no longer BLUE
Some Causes of Why Variance of the Error Will Vary:
- Misspecifications- Omitted Variables
- Measurement Error
2
o
ECON 2003Y (3) Principles of Econometrics
24

Prob > chi2 = 0.3947
chi2(1) = 0.72
Variables: fitted values of dinflation
Ho: Constant variance
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
. hettest
6.2.2. DETECTING HETEROSCEDASTICITY
The Breusch-Pagan LM Test- Linear Model:
H
0
: Homoscedascity: Equal Variance
H
1
: Heteroscedascity: Unequal Variance
Test at 5% significance level with df=10

Critical= 3.94030


Since computed < 2 critical we do not reject the null and conclude that there is significant
evidence of homoscedasticity.
All these tests have shown that our sample is free from these problems. Yet, the regressions are
insignificant. This makes us conclude that neoclassical views today are not representative of the
earlier theory.

7.0. FORECASTING
Phillips curve model can prove useful for forecasting future rates of inflation and has already
been exhaustively explored in the literature.
According to Atkeson and Ohanian (2001) model present day Phillips curve contradicts the trade
off theory. Thus, it does not successfully help to forecast inflation. But instead overestimate the
forecast. We would try to forecast inflation even if it would not be accurate.
According to the Market Watch, predicted average unemployment rate in the U.S for 2014 would
be around 6.8%. If we plug in our regression equation of 2000 to 2013 we obtain expected value
for change in inflation in 2014 which is around -0.11% However, according to Useconomy.com
ECON 2003Y (3) Principles of Econometrics
25

forecasted expected change in inflation is -0.1%. Thus, our Phillips curve model over predicts
the forecasted inflation by about -0.01.


8.0 USING THE MODEL FOR CONTROL OR POLICY PURPOSES.
Let us see the general use of Phillips curve. First, The Central Bank can use Phillips Curve for
the setting of short-term interest rates. Secondly, forecasted inflation affects governments
expected tax revenues. And finally inflation is everywhere and always a monetary phenomenon.
It enables the monetary economists to predict the amount of money, as a proportion of their
income, which people would want to hold at various rates of inflation.
US monetary and fiscal policy in the 1960s were very expansionary implied that policymakers
could exploit the trade off to influence the rate of economic growth and inflation. However, the
relationship from 1970s stated that it was not usable for policy purposes. Once people came to
expect the higher inflation, monetary policy could not keep the unemployment rate permanently
below its natural level.
Classical theory says that any monetary policy in the long run is ineffective as the unemployment
rate will return to its natural rate despite higher rates of inflation. The new classical model
indicates that the effect of expansionary policy depends on whether it is anticipated or
unanticipated. It demonstrates that policy makers cannot know the outcome of their decisions
without knowing the publics expectations regarding them. Once they figure out the publics
expectations, they can know what effect their policies will have. It may be nearly impossible to
find out what the publics expectations are, given that the public consists of over 300 million
U.S. citizens. Public expectations do not remain fixed while policy makers are plotting a
surprisethe public will revise its expectations, and policies will have no predictable effect on
output only unanticipated policy matters.


ECON 2003Y (3) Principles of Econometrics
26

9.0. CONCLUSION
To wrap up our regression analysis of our sample 2000-2013 we had some strength namely no
autocorrelation it had low degree of multicolinearity and equal variance. However, the sample
has unfortunately some weaknesses like low r
2
and insignificant t test and p values.
Nevertheless we have reached at a general conclusion that this particular sample does not exhibit
the presence of Phillips curve.
On a final note this particular sample refutes the backward linear expectations of neoclassical
and could be referred to as misperception theory in modern days. Maybe the data would best suit
the New Keynesian reciprocal approach. Gordon 2013 who followed the new Keynesian
approach found that the Phillips curve is well alive in the U.S.


10.0. REFERENCES
1. Basics Econometrics-Damodar Gujrati 4
th
Edition
2. http://en.wikipedia.org/wiki/Phillips_curve
3. http://www.minneapolisfed.org/research/qr/qr2511.pdf
4. web.uconn.educunninghamecon309phillipscurve.ppt
5. http://ashleymac.econ.vt.edu/working_papers/ashley_verbrugge.pdf

Das könnte Ihnen auch gefallen