# Tarun Singh

**Worked with: Alex Sloan, Alexander Marcus
**

Econometrics Problem Set 1

1. growth

Variable |

Obs

Mean

Std. Dev.

Min

Max

-------------+--------------------------------------------------------------------growth |

65

1.942715

1.89712

-2.811944 7.156855

tradeshr

Variable |

Obs

Mean

Std. Dev.

Min

Max

-------------+--------------------------------------------------------------------tradeshr |

65

.564703

2. Linear regression

.2892703

.140502 1.992616

Number of obs =

F( 1,

63) =

Prob > F

65

12.09

= 0.0009

R-squared

= 0.1237

Root MSE

=

1.79

----------------------------------------------------------------------------------------------|

growth |

Robust

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

**-------------+--------------------------------------------------------------------------------tradeshr | 2.306434 .6632868
**

_cons | .6402653

.4591457

3.48

0.001

.9809608

3.631907

1.39 0.168

-.2772641

1.557795

**----------------------------------------------------------------------------------------------a) The coefficient on tradeshr is 2.306434 meaning as the proportion of GDP
**

that is trade (exports and imports) increases by one unit the average

-2

0

2

4

6

8

**annual percentage growth of real per capita GDP increases by 2.306434
**

units. This value is large since there are more developing countries than

developed countries, and an increase in tradeshr will have a greater effect

on growth in developing countries.

0

.5

1

tradeshr

Fitted values

1.5

2

growth

b)

c) Yes, since the 95% CI for the slope is from .9809608 to 3.631907 and 0 is

not contained in that range the slope is significantly different from 0 at the

5% significance level.

d) .9809608 to 3.631907

**e) The R^2 of the regression is 0.1237. This means that .1237 is the
**

proportion of the sample variance of growth that can be explained by

tradeshr.

f)

Growth

tradeshr

-------------+------------------------------growth | 1.0000

tradeshr | 0.3517 1.0000

The correlation coefficient between growth tradeshr is .3517 which when

squared is .1237. The correlation coefficient is the square root of R^2.

**g) The RMSE is 1.79. This measures the spread of the distribution of the error
**

term.

**h) The regression error appears to be heteroskedastic
**

i)

Non-Robust

Source |

65

SS

df

MS

Number of obs =

-------------+------------------------------

F( 1,

63) =

Model | 28.4885066

0.0041

1

28.4885066

Prob > F

**Residual | 201.851551
**

0.1237

63

3.20399287

R-squared

-------------+-----------------------------Total | 230.340057

64

8.89

=

=

**Adj R-squared = 0.1098
**

3.5990634

Root MSE

=

1.79

-----------------------------------------------------------------------------growth |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------tradeshr | 2.306434

_cons | .6402653

1.619405

.773485

2.98 0.004

.7607473

.4899767

1.31 0.196

-.3388749

3.85212

**In the tables above we see that taking away the “robust” option changed
**

the std. error which also means that the t-values, p-values, and

Confidence Intervals are different as well. The new regression has a higher

Std. Err., lower t value, higher p value, and the confidence interval is

wider.

j) Linear regression

Number of obs =

64

F( 1,

62) =

Prob > F

3.77

= 0.0567

R-squared

= 0.0447

Root MSE

= 1.7894

-----------------------------------------------------------------------------|

growth |

Robust

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

**-------------+---------------------------------------------------------------tradeshr | 1.680905 .8656171
**

3.411249

_cons | .9574107 .5360579

2.028975

1.94

0.057

-.0494392

0.079

-.1141537

1.79

**Yes it does make a qualitative difference, because the R^2 decreased
**

when we removed the outlier meaning that a smaller proportion of the

variance can be explained by the regression compared to the original

regression in part a).

**k) The outlying observation is Malta, an island nation whose economy is
**

largely dependent upon exports and imports thus the tradeshr has a

greater effect on the average annual real per capita GDP growth

compared to a country which is not an island country. The fact that R^2

went down by removing Malta shows that the data from Malta was

skewing the rest of the regression and therefore Malta should be omitted

from the data.

3. a)

Linear regression

Number of obs =

F( 1,

63) =

Prob > F

R-squared

4.36

= 0.0410

= 0.1000

65

Root MSE

= 1.814

-----------------------------------------------------------------------------|

growth |

Robust

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

**-------------+---------------------------------------------------------------lorgdp60 | -1.382122 .6622912
**

_cons |

2.28293 .2176277

-2.09 0.041

10.49 0.000

-2.705605 -.0586384

1.848036

2.717824

**The coefficient on lorgdp60 is -1.382122. This means that the countries in
**

the bottom quartile can expect to have a 1.382122% less growth than the

countries in the other quartiles. This value appears to be large.

**b)Since the 95% Confidence Interval does not have 0 in it this means that at
**

the 5% significance level there appears to be a different mean growth rate for

countries with lorgdp60=0 and countries with lorgdp60=1

**c) sample average for lorgdp60=1 is .9008083
**

sample average for lorgdp60=0 is 2.28293

The difference of means t-statistic is -2.09 and the p value is .041 meaning

the difference significantly different from 0.

**d) The difference of means t-statistic without robust is -2.65. The t statistic
**

computed in c) is -2.09 where as this one is -2.65. this difference is in tstatistic values is due to the non robust regression assuming that the data is

homoskedastic, and due to the robust regression assuming that the data is

heteroskedastic.