Stata Technical Bulletin
31
Two dummy variables are also included to account for the credit controls of 1980 and for an episode in 1973 where price controls
temporarily held the bank loan rate below the commercial paper rate.
Becketti and Morris find that some of the variables in the model do not enter the error correction term, that is, some of the
<5j are zero. They estimate a constrained cointegrating vector using the constrained maximum likelihood procedure of Johansen
and Juselius (1990). In the output for the Hansen test listed below, L.ghat is the lagged value of the estimated error correction
term.
|
. hansen ci finr invb cash rff rmort rtb3, lags(2,2,l. |
,1,1,2,1) static(L.ghat D731 Dec) regress | |||||
|
Quarterly data: 1955:2 |
to 1992:3 |
(160 obs) | ||||
|
Source I |
SS |
df |
MS |
Number of obs F( 13, 136) |
= 160 = 23.02 | |
|
— | ||||||
|
Model I |
.04111377 |
13 .003162698 |
Prob > F |
= 0.0000 | ||
|
Residual ∣ |
.018686145 |
136 .000137398 |
R-square |
= 0.6876 | ||
|
Adj R-square |
= 0.6677 | |||||
|
— | ||||||
|
Total I |
.059799915 |
149 .000401342 |
Root MSE |
= .01172 | ||
|
— ci I |
Coef. |
Std. Err. |
t |
p>∣t∣ |
[967. Conf. |
— Interval] |
|
— | ||||||
|
L.ci I |
.5962585 |
.0811896 |
7.344 |
0.000 |
.4367012 |
.7668169 |
|
L2.ci I |
-.0936603 |
.0791863 |
-1.183 |
0.239 |
-.2602642 |
.0629336 |
|
L.finr I |
.1191324 |
.0686116 |
2.036 |
0.044 |
.0034221 |
.2348428 |
|
L2.finr I |
.0383112 |
.0669799 |
0.672 |
0.602 |
-.0743699 |
.1609923 |
|
L.invb I |
.1985529 |
.1010209 |
1.966 |
0.061 |
-.0012221 |
.398328 |
|
L. cash I |
-.0056775 |
.0333961 |
-0.170 |
0.866 |
-.0717182 |
.0603633 |
|
L.rff I |
-.4328891 |
.2368728 |
-1.836 |
0.069 |
-.899342 |
.0336638 |
|
L.rmort I |
.164ББЗБ |
.2692306 |
0.636 |
0.627 |
-.3480907 |
.6771977 |
|
L2. rmor I |
-.1926119 |
.2218472 |
-0.868 |
0.387 |
-.6313283 |
.2461046 |
|
L.rtb3 I |
.4316918 |
.316006 |
1.366 |
0.174 |
-.1933273 |
1.066611 |
|
L.ghat I |
.0002638 |
.0008643 |
0.297 |
0.767 |
-.0014366 |
.0019433 |
|
D731 I |
.0770014 |
.0120174 |
6.407 |
0.000 |
.0632362 |
.1007666 |
|
Dcc I |
-.0340361 |
.0129617 |
-2.628 |
0.010 |
-.0696489 |
-.0084232 |
|
_cons I |
.003838 |
.0026682 |
1.444 |
0.161 |
-.0014188 |
.0090947 |
Individual Statistics
L.ci = .20142736
L2.ci = .15381964
L.finr = .19617365
L2.finr = .2406588
L.invb = .09370907
L.cash = .31049186
L.rff = .05287014
L.rmort = .0242646
L2.rmor = .0490725
L.rtb3 = .0538896
L.ghat = .38104913
D731 = .52666667
Dcc = .33333333
_cons = .33570572
sigma = .1079207
Model test statistic with 15 degrees of freedom:
_Lc = 2.8214405
The Hansen statistics for changes in the parameters of the reduced form bank loan equation show virtually no evidence of
parameter instability. Among the test statistics for changes in the individual parameters, only the statistic for the error correction
term, L.ghat, is statistically significant at the 10-percent level and only the statistic for the 1973:Q1 dummy, D731, is significant
at the 5-percent level. The joint test for a general breakdown in the equation is not statistically significant. The 5 and 10 percent
critical values with 15 degrees of freedom are 3.54 and 3.26, respectively. Thus the Hansen test supports the conclusion of
Becketti and Morris that the reduced form equation for C&I loans provides little evidence of a change in the elasticity of demand
for bank loans.
Caveat
In addition to tsfit, hansen also utilizes a program called mkmat which stores a variables as an _N × 1 matrix, that is,
as a column vector (Heinecke 1994). Matrices in Stata are constrained by the matsize, a parameter that has a limit of 400 in
the Unix and Intercooled versions of Stata ([5u] matsize). Thus, in its current form, hansen cannot accommodate a model with
more than 399 observations.