Transcript Crosssectional_estimation_STATA

Cross-sectional estimation in STATA
by Binam Ghimire
1
Learning Objectives
 Cross sectional regression in Stata
 Wald test for joint significance
2
Theory
 Many papers have provided evidence of a positive
relationship between financial development and
economic growth.
 Theses papers have mainly applied cross sectional
estimation (average effect) to empirically prove the
above relationship
3
Theory
 Cross-sectional estimation: Beck et al. (2004) paper
shows positive relationship between banks, stock market
and economic growth.
 Next slide shows the output of cross sectional estimation
in Beck et al. (p 430, 2004)
4
Learning Objectives
 Cross sectional regression in Stata
 Wald test for joint significance
5
File to import for practice
 Data from Beck et al. (2004)
 Download the data from
www.banksandmarkets.wordpress.com
 You can note that the name of the countries have been
converted into numbers. (Stata wont recognise the
names therefore you must convert them into numbers).
6
Importing the file
 To import the data in stata: copy and paste from excel
into Data Editor as follows
Data - Data Editor and paste the copied cells. You
should have copied all the cells inside the worksheet
“main” i.e. all 40 observations for all 9 variables
including the country column)
(Or you can import the file using various options
available in Stata: File - Import)
7
Converting data into natural logs
 Generate the natural log of the variables to avoid the
affect of the outliers using the “gen” command as
follows
In the stata command box type
gen lnvariablename =log(variablename)
for some variables you should apply
gen lnvariablename=log(1+variablename)
Press enter each time a gen command is written.
You should convert 8 variables into natural log so 8
times gen command to be applied as follows
 gen lnstart=log(start), gen lnschool=log(1+school) gen
lngov=log(gov), gen lntrade=log(trade), gen lnpi=log(1+pi), gen
lnbmp=log(1+bmp), gen lnpriv=log(priv), gen lntor=log(tor)
8
Converting data into natural logs
 If you make any mistake while generating natural logs
you can drop the generated variable by using the drop
command – e.g. drop lnstart to delete the generated
lnstart.
 You can generate the variable again as shown in
previous slide using the gen command
9
Pooled cross sectional regression
 Now run the cross-sectional pooled regression estimates
using the regress command:
regress growth lnstart lnschool lnpriv lntor
 It should look like the output in next slide. This is the
output of regression 1, simple conditioning information
set of Beck et al. (2004) where the authors have
included the initial real GDP per capita to control for
convergence and the average years of schooling to
control for human capital accumulation.
10
Pooled cross sectional regression
11
Importing Data from Excel into Stata
2. Pooled estimate
 In the policy conditioning information set, authors use
the simple conditioning information set plus either (i)
the black market premium, (ii) the share of exports and
imports to GDP, (iii) the inflation rate or (iv) the ratio of
government expenditures to GDP.
 The regression command for each of the above
regressions are as follows :
regress
regress
regress
regress
growth
growth
growth
growth
lnstart
lnstart
lnstart
lnstart
lnschool
lnschool
lnschool
lnschool
lngov lnpriv lntor
lntrade lnpriv lntor
lnpi lnpriv lntor
lnbmp lnpriv lntor
 The outputs are similar to regression 2, 3, 4 and 5 of
Beck et al. (2004). They are provided in the next 4
slides.
12
Pooled cross sectional regression,
Beck et al. (2004) regression 2
13
Pooled cross sectional regression,
Beck et al. (2004) regression 3
14
Pooled cross sectional regression,
Beck et al. (2004) regression 4
15
Pooled cross sectional regression,
Beck et al. (2004) regression 5
16
Result:
 The OLS regressions demonstrate a strong positive
association between stock market development, bank
development, and economic growth. Both bank
development (bank credit represented by lnpriv) and
stock market development (turnover ratio represented
by lntor) enter each of the five regressions significantly
at the 0.05 significance level.
17
Wald test:
 Wald Test can be used to test the joint significance of a
subset of coefficients.
 Here we can test the two variables from Beck et al.
(2004): lnpriv and lntor.
 The null hypothesis is H0: θ is 0 (Ho: b1=b2=0) and
hence the variables are not providing meaningful
results. The alternative hypothesis is H0: θ is greater or
less than 1 (H1: b1=b2 ≠0) , which means the variables
are jointly significant and give a meaningful result.
 In the regression result lnpriv and lntor are individually
significant based on t-tests with very low p values. But
to test that they are jointly significant we can perform
the Wald test
18
Wald test:
 The command is
Test lnpriv lntor
 The result is shown in the slide next
19
Wald test:
20
Wald test:
 F static at 17.67 with p value of 0.0000 means they are
jointly significant
 The F static of 10.15 and p value of 0.0000 shown in the
regression output is also the indicator of the joint
significance for ALL independent variables. This means
when we perform Wald test for all independent variables
in the first regression equation, the output would be
similar to the F static in OLS equation above (F= 10.15
and = 0.0000). See next slide to check this
21
Wald test:
22
Thank You
23
Reference(s):
 BECK, T. and LEVINE, R. 2004. Stock Markets, Banks,
and Growth: Panel Evidence. Journal of Banking &
Finance, 28, 3, 423-442.
 GUJARATI, D. N. 2003. Basic Econometrics, New York,
McGraw-Hill Education.
24