+ All Categories
Home > Documents > The Eco No Metrics Project

The Eco No Metrics Project

Date post: 07-Apr-2018
Category:
Upload: lice
View: 218 times
Download: 0 times
Share this document with a friend
22
1 | P a g e  Candidate Number 2864B Economics Tripos Part IIA Paper 3  Take Home Examination Question 1: “Do large government deficits raise long-term interest rates?  Assess using time-series data on one country.”  Contents 1. Introduction 2. Theory 3. Literature review 4. The model 5. Data 6. Estimation 7. Conclusion 8. Appendix Word Count: 1993 (Excluding footnotes and Appendix)
Transcript
Page 1: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 1/21

1 | P a g e  

Candidate Number 2864B

Economics Tripos Part IIA Paper 3 

Take Home Examination 

Question 1: “Do large government deficits raise long-term interest rates? 

 Assess using time-series data on one country.”  

Contents

1. Introduction

2. Theory

3. Literature review

4. The model

5. Data

6. Estimation

7. Conclusion

8. Appendix

Word Count: 1993 (Excluding footnotes and Appendix)

Page 2: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 2/21

2 | P a g e  

1. Introduction

The relationship between government deficits and long-term interest rates is a topic of 

frequent debate. The two graphs below show the development of 10 Year US Treasury

Note Yields and the US federal deficit as share of GDP over the last 55 Years.

At first sight, it looks as if there is a weak inverse relationship between the two series, so

that higher deficits are correlated with higher interest rates. Using time series data onthe United States, I will estimate a vector autoregressive model (VAR) to see whether

large government deficits actually raise long-term interest rates.

Page 3: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 3/21

3 | P a g e  

2. Theory

Economic theory suggests several ways in which government deficits can raise long-term

interest rates.

1.  More bonds issued by the government due to higher deficits increases bond

supply, which lowers their price. As bond prices and interest rates are inversely

related, this raises bond interest rates.1 

2.  Persistent government deficits lead to an increasing stock of debt, which could let

investors doubt the long-run ability of the government to serve the debt. They

may demand a higher risk premium, leading to higher interest rates as a result.2 

3.  If consumers viewed increased bond holdings as a consequence of government

deficits as wealth, higher consumption would increase output and money demand,

which subsequently raises the interest rate.3 

4.  In the IS-LM model, a bond-financed fiscal expansion requires an increase in the

interest rate to restore equilibrium in the money market.4 As the government

deficit persists, this will impact on long-term rates.

However, advocates of the Ricardian equivalence hypothesis (REH) endorse a different

view concerning the effect of deficits. According to REH, individuals realize that

government deficits mean future tax increases as the debt must ultimately be paid for,

so they adjust their saving behaviour accordingly. With private saving going up in

response to a decrease in public saving, any crowding out effect is eliminated.

5

 

In addition, causality between deficits and interest rates may run both ways. Reverse

causality could happen through the following channels:6 

1.  Higher interest rates would mean higher servicing costs for the existing stock of 

debt, which increases future deficits.

2.  An increase in interest rates reduces investment, which lowers output and the

capital stock, thereby increasing both the cyclical and structural deficit.

3.  The collapse in investment and subsequent fall in output following higher interest

rates could induce the government to undertake additional investments to

stimulate the economy, thereby raising the deficit even further.

1 ”The Economics of Money, Banking and Financial Markets”, Frederic S. Mishkin, Ch5, 8th Edition.

2 Truman (2001)

3 Ussher (1998), p2.

4 Evans (1987), p282.

5 Barro (1979), p940.

6 Ussher (1998), p13-14.

Page 4: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 4/21

4 | P a g e  

3. Literature Review

Empirical studies examining the relationship between deficits and interest rates have

provided mixed results. Studies that have found positive effects of deficits on interestrates include Hoelscher (1986), Miller and Russek (1991, 1996), Cebula and Koch (1994)

and Engen and Hubbard (2004). Others have found no significant relationships, such as

Zimmerman (1997) or Evans (1985). By 2003, Gale and Orszag (2003) count 30 studies

for the US that find a positive relationship and 30 that do not.

A more recent approach has been the inclusion of “expected deficits” into the analysis.

Studies that incorporate expectations tend to find more significant positive relationships,

such as Elmendorf (1993), who concludes that higher deficits have a positive impact on

five-year bond yields. Laubach (2003) and Laubach, Engen and Hubbard (2004) use CBO

projections to find that increases in projected deficits as well as a higher projected debt-

to-GDP ratio raise long-term interest rates.

Studies using VARs to determine the relationship between deficits and interest rates

include Plosser (1987), Evans (1987), Miller and Russek (1996) and Dai and Phillippon

(2004). Whereas the former two find no relationship between deficits and interest rates,

the study by Miller and Russek concludes that innovations in the deficit explains between

10-50% of the innovations in the long-term interest rate, if Ricardian equivalence

specifications are excluded. Dai and Phillipon use a structural VAR with a no-arbitrage

restriction for their analysis and conclude that a 1% increase in the deficit/GDP ratioraises the 10-Year US Bond yield by 41 basis points.

4. The model

As mentioned earlier, causation between deficits and interest rates may run both ways,

which means that we would have to model both variables as endogenous. If we ignored

this bilateral causality and took government deficits as exogenous to do a single

equation regression, we would have simultaneity bias. The independent variable would

be correlated with the error terms, which renders all OLS estimates biased and

inconsistent.7 

Because of the possible endogeneity, a vector autoregressive model (VAR) was used. A

VAR does not need an a priori distinction between exogenous and endogenous variables.

By using such “atheoretical” VARs, which include all variables as endogenous, we do not

have to impose any prior restrictions. Since all regressors are lagged variables, we can

7 Modern Econometrics, Thomas (1997), Ch8.

Page 5: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 5/21

5 | P a g e  

assume that they are contemporaneously uncorrelated with the error term, so that each

equation can be consistently estimated.8 

In deciding which other variables determine the interest rate and should be included in

the VAR, basic economic theory was applied. According to the theory of liquidity

preference, money demand depends positively on income and negatively on the interest

rate. Together with real money supply, this equality determines the interest rate:9 

Therefore, we should include real GDP, money supply and a measure of the price level as

additional variables. A relationship between nominal and real interest rates is given by

the Fisher equation:10 

This suggests that we should include expected inflation as well if we use nominal interest

rates as dependent variable. Inflation expectations were modelled as adaptive, with

expected inflation being equal to last periods’ inflation. As a VAR includes lags, using

actual inflation will then account for both changes in the price level (as suggested by

equation (1)) and expected inflation. In order to see whether large deficits have a

separate effect on interest rates, a dummy variable for large deficits is included. The

dummy is equal to one if the deficit is higher than its mean value.

Accounting for non-stationarity of some series and seasonal effects, the VAR model then

looks as follows:11 

If higher government deficits raise interest rates, we expect the coefficients γ(1i) to be

positive. If larger deficits have a separate impact on interest rates, we would expect η(1)

to be significant.

8 Modern Econometrics, Thomas (1997), p459.

9 Macroeconomics, Mankiw (2006), Ch4.

10 Macroeconomics, Mankiw (2006), Ch4.

11 ∆i is the change in nominal interest rates, ∆def is the change in deficit (with a positive number denoting a

deficit and a negative number denoting a surplus), ∆rgdp is the growth in real GDP,π

is the inflation rate,∆m2growth is the change in the growth of M2, lar is a dummy for large deficits and q2/3/4 are dummy

variables for the second, third and forth quarter, respectively.

)2(er i π +=

)1(),( /  Y i LP M  =

 

 

 

 

+

 

 

 

 

 

 

 

 

+

 

 

 

 

 

 

 

 

+

 

 

 

 

=

 

 

 

 

=

it 

it 

it 

it 

it 

n

i

iiiii

iiiii

iiiii

iiiii

iiiii

q

q

q

lar 

growm

rdgp

def 

i

growm

rdgp

def 

i

5

4

3

2

1

5453525

4443424

3433323

2423222

1413121

1

55555

44444

33333

22222

11111

5

4

3

2

1

4

3

2

22 ε 

ε 

ε 

ε 

ε 

ω ω ω η 

ω ω ω η 

ω ω ω η 

ω ω ω η 

ω ω ω η 

π 

θ ϕ δ γ   β 

θ ϕ δ γ   β 

θ ϕ δ γ   β 

θ ϕ δ γ   β 

θ ϕ δ γ   β 

α 

α 

α 

α 

α 

π 

Page 6: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 6/21

6 | P a g e  

5. Data12 

The dataset was taken for the United States from the Federal Reserve Economic Data

(FRED). In this context, the 10-Year Treasury Note Rate is used for long-term interest

rates. The time period examined goes from 1966Q2 to 2009Q4. 

The US was chosen for the analysis because it has the most comprehensive dataset with

a large sample size. This is especially important in using VARs because of two reasons:

Firstly, the occurrence of lagged dependent variables in VARs renders the OLS estimates

biased, but they are still consistent.13 Secondly, VARs tend to consume numerous

degrees of freedom because of the large number of regressors due to many lags.

Therefore, it is desirable to have a large sample size in order to obtain precise estimates.

6. Estimation

First, the augmented Dickey-Fuller test was carried out to test for the stationarity of the

series. Non-stationary series (interest rates, deficits, money supply growth) were

differenced, giving rise to a percentage point interpretation of the differenced series. As

the data for the federal deficit is not seasonally adjusted, seasonal dummies were

included to remove seasonal effects. 

In deciding the number of lags to include in the VAR, the Akaike information criterion

suggested a lag length of 3 quarters. However, the LM test for serial correlation indicates

autocorrelation at this lag length. Therefore, successive lags are added until the VAR

model passes all diagnostic tests at the 5% level, giving us 7 lags.14 

The numerical result for this VAR estimation can be found in the Appendix.15 For the

regression with ∆i (changes in the 10-Year Bond rate) as dependent variable, the large 

deficits dummy coefficient is insignificant. The coefficients on ∆def (changes in the

federal deficit) have mixed signs, with the first three lags having negative coefficients

between -0.10 and 0 and the four coefficients after that being positive with values

between 0.06 and 0.08. They are all individually statistically insignificant at the 5% level.

12 See Appendix A3.1 for data source documentation and A3.2 for a full table.

13 Modern Econometrics, Thomas (1997), p209.

14 Diagnostic tests for A1.1.1 VAR_01_66to09: LM autocorrelation test (see Appendix A1.1.2), White

heteroskedasticity test (the joint test gave a p-value of 0.163), Jarque Bera normality test (no normality, but

the large sample size allows the application of the central limit theorem, so the residuals are asymptotically

normally distributed), Ramsey RESET Test (using a squared term for the equation with ∆i as dependent

variable, we obtained a p-value of 0.1993).

15 See Appendix A1.1.1 VAR_01_66to09 for VAR estimation output. See A1.1.2 for residuals plot and

A1.1.3 for impulse functions.

Page 7: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 7/21

7 | P a g e  

However, the coefficients could still be jointly significant. We will test whether all seven

lags of ∆def are relevant in explaining ∆i by using an F-Test for joint exclusion

restrictions. This is testing for Granger causality.

Our null hypothesis is that all the γ(1i) coefficient are jointly zero, which would mean

that ∆def does not Granger cause ∆i. The test shows us that this null hypothesis cannot

be rejected at a 5% significance level.16 Moreover, the Granger test could not detect any

reverse causality running from interest rates to deficits either. We conclude from this

VAR that there is no evidence that higher government deficits raise long-term interest

rates. The only variable that Granger causes (at a 5% level) changes in the interest

rates in this VAR is inflation, which has a positive effect.

By undertaking sensitivity analysis, we will now see whether the negative results are

robust to changing samples and adding variables.

In re-estimating the VAR for the subperiod 1991Q1-2009Q4, we can see whether the

results hold in a low-inflation environment, in which movements in the nominal interest

rate are less influenced by high and volatile inflation.17 The Granger causality test again

finds no Granger causality in any direction, with the large deficit dummy still being

insignificant. 18 Moreover, inflation does not Granger cause changes in the interest rate

any more, suggesting that its significant effect was confined to the high inflation period

in the 70s and 80s.

Now we include a new variable to our VAR, the current account as share of GDP. Possible

effects of deficits on interest rates could be diminished by the fact that financial markets

have become increasingly integrated. High deficits then do not necessarily increase

interest rates via the crowding out effect, as the government can borrow funds from

abroad. To account for the effect of capital flows, the current account as share of GDP

(differenced for stationarity) was included as another endogenous variable.

The modified VAR was estimated for the time period 1966Q2-2009Q4 using 7 lags and

passes all the diagnostic tests.19 It yields the same qualitative results as our first VAR.20 

The γ(1i) coefficients are all individually insignificant and range between -0.17 and 0.12.

16 See Appendix A1.1.4 VAR_01_66to09 Granger causality test 

17 Diagnostic tests for A1.2.1 VAR_02_91to09: LM autocorrelation test (p-values for 8 lags ranged from

0.2801 to 0.9687, with a p-value of 0.03 for the 8th lag) White heteroskedasticity test (the p-value of joint test

is 0.4125), Jarque Bera normality test (no normality, but the large sample size allows the use of central limit

theorem, so residuals are asymptotically normally distributed), Ramsey RESET Test (using a squared term for

the equation with ∆i as dependent variable, we obtained a p-value of 0.8655).

18 See Appendix A1.2.1 VAR_02_91to09 for estimation output and A1.2.2 VAR_02_91to09 Granger

causality test 

19 Diagnostic tests: LM autocorrelation test (the p-values for 8 lags ranged from 0.0563 to 0.4989), White

heteroskedasticity test (the p-value of the joint test is 0.2279), Jarque Bera normality test (no normality, but

large sample size allows use of central limit theorem), Ramsey RESET Test (using a squared term for the

equation with ∆i as dependent variable, we obtained a p-value of 0.2528).

20 See Appendix A1.3.1 VAR_03_CurrentAccount for VAR estimation output

Page 8: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 8/21

8 | P a g e  

A test for Granger causality shows that there is no evidence for past changes in deficits

Granger causing changes in the interest rates at a 5% significance level.21 The coefficient

on the large deficit dummy is insignificant. Again, inflation is the only variable that

Granger-causes changes in the interest rate.

Sensitivity analysis has supported the robustness of our results. We conclude that there

is no evidence for government deficits of any size raising long-term interest rates.

7. Conclusion

Using a vector autoregressive model, we found no evidence for the claim that large

government deficits raise long-term interest rates. This does not necessarily prove

Ricardian equivalence right, since the actual mechanism through which interest rates are

equilibrated may be quite different from what REH postulates. One possible explanation

for our negative results is that worldwide capital flows were insufficiently modelled. One

way to deal with this is to estimate the model for the world as a closed economy, where

a first attempt has been made by Ford and Laxton (1999). Possible effects could also be

mitigated for the US in particular, because US bonds are considered as a very safe asset.

Investors might be less worried about US budget deficits than they would be in other

countries, which reduces the risk premium effect. Part of the results can also be

attributed to shortcomings of the VAR model, as VARs are more likely to suffer from

measurement error, which biases the coefficients towards zero.

22

In the future, moreefforts can be undertaken in the direction of looking at expected deficits rather than past

deficits, where studies have already delivered promising results. Taken together, all

these suggestions and limitations point to much scope for future studies in this area of 

research. 

21 See Appendix A1.3.2 VAR_03_CurrentAccount Granger causality test 

22 Gale and Orszag (2004), p27.

Page 9: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 9/21

9 | P a g e  

8. Appendix

Overview

A1.1.1 Table: Vector autoregression estimation output VAR_01_66to09

A1.1.2 Graph/Table: VAR_01_66to09 residuals plot and serial correlation LM Test

A1.1.3 Graph: VAR_01_66to09 Impulse response functions

A1.1.4 Table: VAR_01_66to09 Granger causality test results

A1.2.1 Table: Vector autoregression estimation output VAR_02_91to09

A1.2.2 Table: VAR_02_91to09 Granger causality test results

A1.3.1 Table: Vector autoregression estimation output VAR_03_CurrentAccount

A1.3.2 Table: VAR_03_CurrentAccount Granger causality test results

A2 References

A3.1 Data source documentation

A3.2 Table: Full Data table

Page 10: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 10/21

10 | P a g e  

A1.1.1 Table: Vector autoregression estimation output VAR_01_66to09 

VAR_01_66to09 Included observations: 167 after adjustments

Sample (adjusted): 1968Q2 2009Q4 Standard errors in ( ) & t-statistics in [ ]

DINTEREST DDEFICIT GDPGROWTH INFLATION DM2GROWTH

C -0.265141 -0.171103 0.594538 -0.173569 0.481903

-0.18616 -0.17026 -0.30275 -0.23682 -0.29615

[-1.42430] [-1.00498] [ 1.96377] [-0.73293] [ 1.62723]

DDEFICIT(-1) -0.001414 -0.709845 -0.258683 -0.167765 -0.213776

-0.08925 -0.08163 -0.14516 -0.11354 -0.14199

[-0.01584] [-8.69581] [-1.78207] [-1.47753] [-1.50555]

DDEFICIT(-2) -0.062026 -0.533681 -0.18734 0.117582 -0.307614

-0.11399 -0.10425 -0.18538 -0.14501 -0.18134

[-0.54415] [-5.11918] [-1.01055] [ 0.81086] [-1.69635]

DDEFICIT(-3) -0.102381 -0.426973 -0.105671 0.19768 -0.265578

-0.12142 -0.11105 -0.19747 -0.15447 -0.19317

[-0.84319] [-3.84486] [-0.53512] [ 1.27977] [-1.37487]

DDEFICIT(-4) 0.07437 -0.084439 -0.191795 0.22464 -0.381472

-0.12801 -0.11708 -0.20819 -0.16285 -0.20365

[ 0.58096] [-0.72121] [-0.92124] [ 1.37942] [-1.87317]

DDEFICIT(-5) 0.077708 -0.209783 0.079429 0.318017 -0.478322-0.12204 -0.11162 -0.19848 -0.15526 -0.19415

[ 0.63673] [-1.87946] [ 0.40018] [ 2.04833] [-2.46362]

DDEFICIT(-6) 0.063933 -0.116505 -0.269467 0.12596 -0.380494

-0.12128 -0.11092 -0.19725 -0.15429 -0.19294

[ 0.52714] [-1.05031] [-1.36613] [ 0.81639] [-1.97204]

DDEFICIT(-7) 0.072754 0.000871 -0.173464 0.021897 -0.107968

-0.0953 -0.08716 -0.15499 -0.12123 -0.1516

[ 0.76345] [ 0.00999] [-1.11923] [ 0.18062] [-0.71217]

DUMMY_LARGE -0.101984 0.558235 -0.116876 -0.19461 0.08613

-0.09378 -0.08577 -0.15251 -0.1193 -0.14919

[-1.08753] [ 6.50879] [-0.76633] [-1.63131] [ 0.57734]

… other variables

R-squared 0.376305 0.734448 0.372323 0.61561 0.426628

Adj. R-squared 0.184776 0.652901 0.179571 0.497569 0.250553

Sum sq. resids 28.90378 24.17714 76.45097 46.77675 73.15163

Page 11: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 11/21

11 | P a g e  

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

70 75 80 85 90 95 00 05

DINTEREST Residuals

-2

-1

0

1

2

3

70 75 80 85 90 95 00 05

DDEFICIT Residuals

-2

-1

0

1

2

3

70 75 80 85 90 95 00 05

GDPGROWTH Residuals

-3

-2

-1

0

1

2

70 75 80 85 90 95 00 05

INFLATION Residuals

-2

-1

0

1

2

70 75 80 85 90 95 00 05

DM2GROWTH Residuals

A1.1.2  Graph/Table: VAR_01_66to09 residuals plot and serial correlation

LM Test 

Page 12: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 12/21

12 | P a g e  

A1.1.3 Graph: VAR_01_66to09 Impulse response functions 

-.20

-.15

-.10

-.05

.00

.05

.10

.15

.20

1 2 3 4 5 6 7 8 9 10

Response of DDEFICIT to CholeskyOne S.D. DINTEREST Innovation

-.15

-.10

-.05

.00

.05

.10

.15

1 2 3 4 5 6 7 8 9 10

Response of DINTEREST to Cholesky

One S.D. DDEFICIT Innovation

Page 13: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 13/21

13 | P a g e  

A1.1.4  Table: VAR_01_66to09 Granger causality test results 

VAR Granger Causality/Block Exogeneity Wald Tests

Date: 05/05/10 Time: 19:27

Sample: 1947Q1 2009Q4

Included observations: 167

Dependent variable: DINTEREST

Excluded Chi-sq df Prob.

DDEFICIT 3.643830 7 0.8198

GDPGROWTH 8.104320 7 0.3235

INFLATION 17.09830 7 0.0168

DM2GROWTH 11.91664 7 0.1033

All 39.99320 28 0.0662

Dependent variable: DDEFICIT

Excluded Chi-sq df Prob.

DINTEREST 10.77719 7 0.1486

GDPGROWTH 12.62192 7 0.0819

INFLATION 6.525555 7 0.4799

DM2GROWTH 7.561830 7 0.3728

All 46.79342 28 0.0144

Page 14: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 14/21

14 | P a g e  

A1.2.1 Table: Vector autoregression estimation output VAR_02_91to09 

VAR_02_91to09 Included observations: 76

Sample: 1991Q1 2009Q4 Standard errors in ( ) & t-statistics in [ ]

DINTEREST DDEFICIT GDPGROWTH INFLATION DM2GROWTH

C -0.390838 -0.702851 1.446004 1.903676 -0.283247

-0.44715 -0.72929 -0.83798 -0.7474 -0.73623

[-0.87406] [-0.96375] [ 1.72559] [ 2.54706] [-0.38473]

DDEFICIT(-1) -0.090547 -0.625071 -0.23924 -0.286478 0.059454

-0.1227 -0.20013 -0.22995 -0.2051 -0.20203

[-0.73794] [-3.12340] [-1.04040] [-1.39680] [ 0.29428]

DDEFICIT(-2) 0.085836 -0.441449 -0.333808 -0.324128 -0.049441-0.17767 -0.28977 -0.33295 -0.29696 -0.29253

[ 0.48313] [-1.52346] [-1.00257] [-1.09147] [-0.16901]

DDEFICIT(-3) 0.07606 -0.214111 -0.389666 -0.429618 0.013112

-0.19443 -0.31712 -0.36438 -0.32499 -0.32014

[ 0.39119] [-0.67518] [-1.06940] [-1.32193] [ 0.04096]

DDEFICIT(-4) 0.138224 0.118391 -0.293815 -0.557069 -0.147529

-0.19939 -0.32519 -0.37366 -0.33327 -0.32829

[ 0.69325] [ 0.36406] [-0.78632] [-1.67152] [-0.44939]

DDEFICIT(-5) 0.257936 -1.38E-02 -0.230108 -0.268976 -0.427319

-0.17988 -0.29337 -0.3371 -0.30066 -0.29617

[ 1.43397] [-0.04695] [-0.68262] [-0.89462] [-1.44284]

DDEFICIT(-6) 0.141545 0.097133 -0.405803 -0.233567 -0.626154

-0.16356 -0.26676 -0.30652 -0.27339 -0.2693

[ 0.86540] [ 0.36412] [-1.32391] [-0.85434] [-2.32510]

DDEFICIT(-7) 0.03267 0.101211 -0.213772 -0.063592 -0.387133

-0.12343 -0.2013 -0.23131 -0.2063 -0.20322

[ 0.26469] [ 0.50277] [-0.92420] [-0.30824] [-1.90498]

DUMMY_LARGE -0.084003 0.76526 -0.042911 -0.183602 0.212917

-0.13448 -0.21934 -0.25202 -0.22478 -0.22142

[-0.62464] [ 3.48900] [-0.17027] [-0.81680] [ 0.96159]

… other variables

R-squared 0.675231 0.832256 0.559291 0.556533 0.817453

Adj. R-squared 0.323399 0.650534 0.081856 0.076111 0.619693

Sum sq. resids 3.63314 9.664411 12.75973 10.15041 9.849296

Page 15: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 15/21

Page 16: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 16/21

16 | P a g e  

A1.3.1  Table: Vector autoregression estimation output

VAR_03_CurrentAccount 

VAR_03_CurrentAccount Included observations: 167 after adjustments

Sample (adjusted): 1968Q2 2009Q4 Standard errors in ( ) & t-statistics in [ ]

DINTEREST DDEFICIT GDPGROWTH INFLATION DM2GROWTH DCURACNT

C -0.271418 -0.165014 0.615461 -0.219424 0.496245 -0.010857

-0.18432 -0.16976 -0.30074 -0.2292 -0.29803 -0.0316

[-1.47255] [-0.97202] [ 2.04647] [-0.95733] [ 1.66510] [-0.34352]

DDEFICIT(-1) 0.002194 -0.706698 -0.238593 -0.179661 -0.181892 0.031248

-0.08999 -0.08289 -0.14684 -0.11191 -0.14551 -0.01543

[ 0.02438] [-8.52612] [-1.62489] [-1.60545] [-1.25002] [ 2.02505]

DDEFICIT(-2) -0.095524 -0.503992 -0.163557 0.037488 -0.252777 0.026233

-0.11599 -0.10683 -0.18925 -0.14423 -0.18754 -0.01989

[-0.82356] [-4.71772] [-0.86423] [ 0.25991] [-1.34783] [ 1.31900]

DDEFICIT(-3) -0.172273 -0.364954 -0.074756 0.054408 -0.179121 0.051732

-0.12395 -0.11416 -0.20224 -0.15413 -0.20041 -0.02125

[-1.38988] [-3.19686] [-0.36964] [ 0.35300] [-0.89376] [ 2.43413]

DDEFICIT(-4) 0.019203 -0.042325 -0.166438 0.069896 -0.308005 0.0111

-0.1317 -0.1213 -0.21489 -0.16377 -0.21295 -0.02258

[ 0.14581] [-0.34893] [-0.77454] [ 0.42679] [-1.44639] [ 0.49153]

DDEFICIT(-5) 0.019335 -0.13822 0.131495 0.152479 -0.361103 4.95E-05

-0.12514 -0.11525 -0.20418 -0.15561 -0.20234 -0.02146

[ 0.15451] [-1.19926] [ 0.64402] [ 0.97988] [-1.78468] [ 0.00231]

DDEFICIT(-6) 0.028568 -0.085275 -0.217834 0.032243 -0.324293 -0.015448

-0.12232 -0.11266 -0.19958 -0.1521 -0.19778 -0.02097

[ 0.23356] [-0.75694] [-1.09148] [ 0.21198] [-1.63971] [-0.73657]

DDEFICIT(-7) 0.117141 0.049448 -0.175084 -0.025364 -0.04151 -0.021007

-0.09937 -0.09153 -0.16214 -0.12357 -0.16068 -0.01704

[ 1.17881] [ 0.54027] [-1.07983] [-0.20526] [-0.25834] [-1.23288]

DUMMY_LARGE -0.150708 0.59505 -0.073437 -0.282073 0.166227 0.026412-0.09508 -0.08757 -0.15514 -0.11823 -0.15374 -0.0163

[-1.58507] [ 6.79501] [-0.47337] [-2.38573] [ 1.08125] [ 1.62008]

… other variables

R-squared 0.426186 0.752229 0.418756 0.662089 0.455067 0.444215

Adj. R-squared 0.206224 0.65725 0.195946 0.532557 0.246176 0.231164

Sum sq. resids 26.59215 22.55831 70.79543 41.12065 69.52333 0.781835

Page 17: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 17/21

17 | P a g e  

A1.3.2  Table: VAR_03_CurrentAccount Granger causality test results

VAR Granger Causality/Block Exogeneity Wald Tests

Date: 05/05/10 Time: 19:45

Sample: 1947Q1 2009Q4

Included observations: 167

Dependent variable: DINTEREST

Excluded Chi-sq df Prob.

DDEFICIT 5.620068 7 0.5847

GDPGROWTH 8.625229 7 0.2807

INFLATION 17.40420 7 0.0150

DM2GROWTH 10.31800 7 0.1713

DCURACNT 10.43150 7 0.1654

All 51.50530 35 0.0356

Dependent variable: DDEFICIT

Excluded Chi-sq df Prob.

DINTEREST 9.930980 7 0.1925

GDPGROWTH 13.99562 7 0.0513

INFLATION 7.460963 7 0.3825

DM2GROWTH 6.758907 7 0.4544DCURACNT 8.611454 7 0.2818

All 55.99862 35 0.0136

Page 18: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 18/21

18 | P a g e  

A2 References

 “Do Budget Deficits Raise Nominal Interest Rates? Evidence from six countries”, Paul Evans (1987),

Journal of Monetary Economics 20 (1987).

 “Do Budget Deficits Raise Interest Rates?” L. Ussher (1998), p4, Working Papers Department of 

Economics, Queens College of the City University of New York.

 “Budget Deficits, National Saving and Interest Rates”, W. G. Gale and P. R. Orszag (2004), Brookings

Institution and Tax Policy Center.

 “World public debt and real interest rates”, R. Ford and D. Laxton (1999), International Monetary Fund.

 “Federal government debt and interest rates”, Eric M. Engen and R. Glenn Hubbard (2004), NBER

Macroeconomics Annual, Vol.19 (2004).

 “A note on interest rates and structural federal budget deficits.” John Kitchen (2002), MPRA Paper No.

21069.

 “Budget Deficits and Interest Rates. A fresh perspective”, Ari Aisen and David Hauner (2008), IMF

Working Paper.

 “Government debt”, Douglas W. Elmendorf and N. Gregory Mankiw (1998), Handbook of 

Macroeconomics.

 “New evidence on Deficits and Interest Rates.”, Gregory Hoelscher (1986), Journal of Money, Credit and

Banking, Vol.18, No.1 (Feb 1987).

 “New evidence on the Interest Rate effects of Budget Deficits and Debt.” Thomas Laubach (2003).

 “Fiscal policy and the term structure”, Charles Plosser (1982), Elsevier Science Publishers.

 “A note on budget deficits and Interest Rates. Evidence from a small open economy”, George

Vamvoukas (1997), Southern Economic Journal, Vol.63, No.3.

 “A no arbitrage vector autoregression of term structure dynamics with macroeconomic and latent

variables.” Andrew Ang and Monica Piazzesi (2003), Journal of Monetary Economics 50 (2003).

 “Government deficits and interest rates. A no-arbitrage structural VAR approach.” Qiang Dai and Thomas

Phillipon (2004), New York University.

 “Do Federal deficits affect interest rates? Evidence from 3 econometric methods.”, Stephen M. Miller and

Frank S. Russek (1996), Journal of Macroeconomics, Vol.18.

 “A Macro finance model of the term structure, monetary policy and the economy.”, Glenn D. Rudebusch,

Tao Wu (2003).

 “The effects of Budget deficits on interest rates: A review of empirical results.” Thomas Laubach.

Page 19: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 19/21

19 | P a g e  

A2.1 Data source documentation

All data is taken from the Federal Reserve Economic Data (FRED), managed by the

Federal Reserve branch in St. Louis: http://research.stlouisfed.org/fred2/ 

10 Year Treasury Note monthly rates (GS10)

http://research.stlouisfed.org/fred2/series/GS10?cid=115 

For long-term interest rates, the monthly 10-Year Treasury Note rate was used,averaged over three months to obtain quarterly data.

Federal government debt: Total public debt (GFDEBTN), not adjusted 

http://research.stlouisfed.org/fred2/series/GFDEBTN?cid=5 

For the deficit, data on total federal debt in million dollars was used: Each value was

subtracted from the quarter before to get the deficit, then division by nominal GDP was

carried out to obtain the deficit in a quarter as percentage of GDP. In this connection, a

positive number is interpreted as a deficit and a negative number as a surplus.

Nominal GDP (GDP)

Used to calculate the deficit and current account balance, seasonally adjusted: 

http://research.stlouisfed.org/fred2/series/GDP?cid=106 

Real GDP growth (GDPC1), seasonally adjusted

http://research.stlouisfed.org/fred2/series/GDPC1?cid=106 

Quarterly data for real GDP was used to get a measure of quarterly growth.

Page 20: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 20/21

20 | P a g e  

Consumer price index for all urban consumers, all items (CPIAUCSL),

seasonally adjusted

http://research.stlouisfed.org/fred2/series/CPIAUCSL?cid=9 

Inflation data was calculated using the CPI, dividing each value by the value three

months ago in order to get inflation per quarter.

M2 Money stock (M2SL), seasonally adjusted

http://research.stlouisfed.org/fred2/series/M2SL?cid=29 

Money supply growth was obtained by using monthly data on the stock of M2, dividingeach observation by the observation three months ago, which resulted in quarterly M2

growth.

Balance of current account (BOPBCA), seasonally adjusted

http://research.stlouisfed.org/fred2/series/BOPBCA?cid=125 

In dividing the balance on current account by nominal GDP, the current account balanceas a share of GDP was obtained.

A3.2 Full Data table

Page 21: The Eco No Metrics Project

8/6/2019 The Eco No Metrics Project

http://slidepdf.com/reader/full/the-eco-no-metrics-project 21/21

21 | P a g e

Year Quarter

10 Year

Treasury

Deficit %

of GDP

Real GDP

growth Inflation

M2

Growth

CA as %

of GDP Year Quarter

10 Year

Treasury

Deficit %

of GDP

Real GDP

growth Inflation

M2

Growth

CA as %

of GDP

1966 1 4.770 0.333 1.255 0.127 1988 1 8.417 1.104 1.284 1.034 2.223 -0.666

2 4.780 -0.618 0.658 0.527 0.341 0.101 2 8.910 1.169 0.516 1.109 1.279 -0.571

3 5.140 1.072 0.810 1.233 1.019 0.061 3 9.100 1.038 1.335 1.181 0.623 -0.543

4 5.003 0.559 0.881 0.152 1.240 0.096 4 8.957 1.534 0.939 1.084 0.885 -0.598

1967 1 4.583 0.198 0.021 0.608 2.180 0.107 1989 1 9.207 1.036 0.748 1.568 0.477 -0.532

2 4.820 -0.962 0.798 0.906 2.886 0.081 2 8.773 1.067 0.793 1.137 1.547 -0.461

3 5.247 1.525 0.763 0.898 2.350 0.067 3 8.107 1.030 0.218 0.723 2.037 -0.399

4 5.643 0.996 2.061 1.187 1.775 0.056 4 7.907 1.674 1.045 1.675 1.666 -0.426

1968 1 5.610 0.532 1.698 0.880 1.574 0.022 1990 1 8.423 1.707 0.397 1.098 1.100 -0.414

2 5.743 -0.446 0.684 1.453 1.848 0.036 2 8.677 1.569 -0.002 1.241 0.683 -0.337

3 5.460 1.001 0.433 1.146 2.199 0.015 3 8.703 1.532 -0.876 2.222 1.078 -0.3684 5.770 0.342 1.575 1.133 2.098 -0.005 4 8.397 2.236 -0.484 0.975 0.879 -0.244

1969 1 6.177 0.155 0.291 1.681 1.124 0.013 1991 1 8.017 1.683 0.675 0.297 1.373 0.169

2 6.353 -0.668 0.632 1.377 0.660 -0.013 2 8.130 1.207 0.421 0.814 0.689 0.042

3 6.857 0.776 -0.470 1.359 0.673 0.005 3 7.940 2.090 0.392 0.734 0.119 -0.069

4 7.297 0.741 -0.157 1.609 1.063 0.036 4 7.347 2.203 1.098 0.802 0.618 -0.089

1970 1 7.367 0.366 0.181 1.583 -0.204 0.061 1992 1 7.303 1.264 1.063 0.795 0.546 -0.101

2 7.713 -0.182 0.890 1.039 1.818 0.094 2 7.377 1.618 1.032 0.789 -0.179 -0.189

3 7.460 0.815 -1.060 1.285 2.888 0.049 3 6.617 1.231 1.052 0.854 0.886 -0.230

4 6.853 0.954 2.758 1.269 2.693 0.021 4 6.743 1.717 0.184 0.776 -0.128 -0.289

1971 1 6.017 0.224 0.567 0.501 4.013 0.062 1993 1 6.280 0.809 0.640 0.700 -0.222 -0.226

2 6.247 0.495 0.798 1.247 3.220 -0.037 2 5.990 1.815 0.526 0.487 0.899 -0.312

3 6.483 1.300 0.278 0.739 2.766 -0.047 3 5.617 0.874 1.321 0.761 0.432 -0.318

4 5.890 0.997 1.788 0.733 2.763 -0.102 4 5.607 1.796 0.974 0.481 0.540 -0.412

1972 1 6.033 0.262 2.372 0.728 2.884 -0.142 1994 1 6.067 0.570 1.368 0.615 0.282 -0.361

2 6.143 -0.073 0.958 0.723 2.858 -0.138 2 7.083 0.981 0.644 0.815 0.183 -0.406

3 6.290 0.584 1.647 0.957 3.608 -0.103 3 7.333 0.648 1.111 0.674 -0.123 -0.443

4 6.373 1.088 2.558 1.185 2.974 -0.088 4 7.837 1.470 0.245 0.736 0.212 -0.503

1973 1 6.603 0.739 1.157 2.342 1.160 0.011 1995 1 7.483 0.870 0.215 0.864 0.183 -0.430

2 6.807 -0.093 -0.534 1.144 2.050 0.064 2 6.620 1.171 0.840 0.527 1.959 -0.435

3 7.207 0.230 0.954 3.167 0.729 0.197 3 6.323 0.300 0.697 0.590 1.278 -0.3624 6.753 0.585 -0.877 2.632 2.029 0.236 4 5.893 0.192 0.685 0.782 0.958 -0.307

1974 1 7.053 0.310 0.256 2.778 1.535 0.111 1996 1 5.910 1.655 1.729 0.905 1.387 -0.359

2 7.543 0.037 -0.989 2.495 0.974 0.007 2 6.720 0.548 0.870 0.577 1.055 -0.383

3 7.963 0.466 -0.394 3.448 1.350 -0.019 3 6.780 0.794 1.092 0.764 0.852 -0.457

4 7.670 0.714 -1.216 2.549 1.455 0.034 4 6.343 1.209 0.769 0.759 1.379 -0.392

1975 1 7.540 1.059 0.764 1.338 3.178 0.268 1997 1 6.563 0.697 1.483 0.314 1.141 -0.443

2 8.050 1.415 1.684 1.887 4.278 0.312 2 6.697 -0.056 1.255 0.313 1.221 -0.348

3 8.297 1.194 1.306 1.667 2.328 0.245 3 6.243 0.435 0.767 0.686 1.655 -0.389

4 8.063 1.298 2.273 1.639 2.886 0.281 4 5.907 1.038 0.945 0.310 1.676 -0.507

1976 1 7.753 1.321 0.752 0.538 3.331 0.145 1998 1 5.587 0.460 0.899 0.123 1.974 -0.512

2 7.773 1.085 0.490 1.604 2.404 0.091 2 5.597 0.062 1.319 0.617 1.554 -0.590

3 7.730 0.757 0.726 1.579 3.563 -0.005 3 5.203 -0.241 1.731 0.429 2.505 -0.666

4 7.190 0.972 1.161 1.382 3.573 0.009 4 4.670 0.962 0.891 0.488 2.169 -0.674

1977 1 7.353 0.781 1.987 2.215 2.952 -0.139 1999 1 4.983 0.404 0.782 0.729 1.443 -0.692

2 7.370 0.253 1.788 1.333 2.259 -0.151 2 5.540 -0.136 1.272 0.482 1.510 -0.779

3 7.357 1.157 -0.021 1.316 2.225 -0.133 3 5.883 0.182 1.796 0.840 1.231 -0.858

4 7.597 0.935 0.341 1.786 2.049 -0.278 4 6.140 1.234 0.261 0.714 1.666 -0.890

1978 1 8.010 0.836 3.934 1.914 1.618 -0.327 2000 1 6.480 -0.027 1.951 0.945 2.078 -1.018

2 8.320 0.474 0.980 2.504 1.823 -0.166 2 6.177 -0.873 0.084 1.053 0.461 -1.001

3 8.490 0.932 1.323 2.443 2.130 -0.157 3 5.893 -0.116 0.592 0.695 1.757 -1.0804 8.820 0.717 0.167 2.086 1.427 -0.028 4 5.567 -0.118 -0.330 0.978 2.208 -1.094

1979 1 9.107 0.300 0.094 3.066 2.224 -0.017 2001 1 5.050 1.083 0.656 0.456 3.177 -1.058

2 9.113 0.312 0.719 3.399 2.332 -0.027 2 5.270 -0.455 -0.274 0.567 1.483 -0.948

3 9.103 0.812 0.275 3.014 1.784 0.036 3 4.980 0.777 0.353 0.113 2.575 -1.017

4 10.447 0.683 0.322 3.723 1.527 -0.004 4 4.770 1.295 0.859 0.056 2.292 -0.851

1980 1 11.987 0.672 -2.049 3.718 1.315 -0.127 2002 1 5.077 0.590 0.531 0.900 0.744 -0.992

2 10.477 0.509 -0.186 2.101 2.882 -0.034 2 5.100 1.125 0.500 0.390 1.730 -1.090

3 10.953 1.032 1.850 2.542 2.543 0.156 3 4.260 0.945 0.021 0.667 2.010 -1.085

4 12.423 0.738 2.078 2.952 1.394 0.082 4 4.007 1.630 0.405 0.773 1.697 -1.145

1981 1 12.960 1.113 -0.798 2.179 3.255 0.032 2003 1 3.920 0.500 0.798 0.329 1.686 -1.241

2 13.750 0.209 1.215 2.694 1.368 0.040 2 3.620 1.860 1.676 0.273 2.476 -1.184

3 14.847 0.835 -1.246 2.077 2.372 0.066 3 4.233 0.992 0.899 0.653 0.202 -1.156

4 14.087 0.969 -1.641 1.071 2.823 0.024 4 4.287 1.851 0.704 0.757 0.132 -1.103

1982 1 14.293 1.005 0.542 0.636 1.892 -0.009 2004 1 4.020 1.130 0.711 0.590 1.968 -1.178

2 13.930 0.560 -0.386 2.632 1.652 0.117 2 4.600 1.199 0.735 0.907 1.403 -1.325

3 13.117 1.884 0.079 0.615 2.181 -0.123 3 4.303 0.862 0.868 0.899 1.347 -1.331

4 10.667 1.628 1.244 -0.204 4.734 -0.151 4 4.173 1.754 0.998 0.419 0.765 -1.477

1983 1 10.563 1.362 2.248 0.919 3.501 -0.074 2005 1 4.297 1.444 0.426 1.096 0.601 -1.409

2 10.543 2.093 1.972 1.012 1.772 -0.225 2 4.160 0.467 0.760 0.620 1.206 -1.431

3 11.627 1.563 2.067 1.002 1.664 -0.360 3 4.213 0.745 0.517 2.155 1.509 -1.455

4 11.687 0.880 1.940 1.290 1.851 -0.419 4 4.490 1.803 1.312 0.050 1.385 -1.625

1984 1 11.943 1.358 1.727 1.175 2.500 -0.550 2006 1 4.570 1.504 0.360 0.703 1.021 -1.507

2 13.200 1.231 0.972 0.774 1.454 -0.600 2 5.070 0.363 0.027 1.147 1.299 -1.514

3 12.867 1.477 0.814 0.961 1.586 -0.588 3 4.897 0.639 0.731 -0.542 1.731 -1.597

4 11.743 2.203 0.944 0.571 3.264 -0.659 4 4.630 1.256 0.300 0.779 1.725 -1.381

1985 1 11.583 1.144 0.847 1.230 1.837 -0.571 2007 1 4.680 1.211 0.794 1.272 1.466 -1.443

2 10.813 1.501 1.562 0.654 2.279 -0.687 2 4.847 0.127 0.887 0.822 1.273 -1.361

3 10.337 1.122 0.759 0.743 1.587 -0.735 3 4.730 0.976 0.527 0.685 1.811 -1.210

4 9.760 2.803 0.961 1.290 1.384 -0.802 4 4.260 1.541 -0.182 1.508 1.321 -1.153

1986 1 8.557 0.924 0.402 -1.092 2.216 -0.781 2008 1 3.663 1.438 0.362 0.886 2.323 -1.247

2 7.603 1.615 0.964 0.736 2.715 -0.807 2 3.887 0.374 -0.676 2.333 1.138 -1.295

3 7.307 1.452 0.483 0.639 2.331 -0.847 3 3.863 3.713 -1.371 -1.036 2.695 -1.266

4 7.263 1.942 0.554 1.089 2.088 -0.863 4 3.253 4.761 -1.647 -2.247 3.667 -1.079

1987 1 7.193 0.680 1.063 1.167 0.874 -0.852 2009 1 2.737 3.018 -0.185 0.401 0.504 -0.735

2 8.343 1.313 0.867 0.976 0.411 -0.853 2 3.313 2.937 0.555 0.923 1.141 -0.691

3 8.877 0.839 1.711 1.054 1.297 -0.842 3 3.517 2.522 1.360 0.737 0.595 -0.719

4 9.123 1.646 0.517 0.870 1.178 -0.845 4 3.460 2.750 0.800 0.569 -0.027 -0.800


Recommended