Economic Models - Convex Optimization
Economic Models - Convex Optimization Economic Models - Convex Optimization
Credit and Income 213 4.4. Essential Remarks In seminal works about money and income (as for instance in Tobin, 1970), the analysis is restricted in a pure theoretical consideration. Some authors (see for instance, Friedman and Kuttner, 1992) write many explanations about co-integration, co-integrating vectors, and error-correction models, but I did not trace any attempt to compute either. In a later paper by the same authors (Friedman and Kuttner, 1993), I read (p. 200) the phrase, “vector autoregression system,” which in fact is understood as a VAR. The point is that no VAR model is specified in this work. Besides, the lag length is determined by simply applying the F-test (p. 191). In this context, the validity of the results is questionable. In other applications (see for instance, Arestis and Demetriades, 1997), where a VAR model has been formulated, nothing is mentioned about the test applied to determine the maximum lag length. Also, in this paper, the authors are restricted to the use of trace statistic (l (trace) ) to test for co-integration (p. 787), although this is the necessary but not the sufficient condition, as pointed out above. From other research works (see for instance, Arestis et al., 2001), I get the impression that there is not a clear notion regarding stationarity and/or integration, since the authors state [p. 22 immediately after Eq. (1)], that “D isaset of I(0) deterministic variables such as constant, trend and dummies. . . ”. It is clear that this verification is false. How can a trend, for instance, be a stationary, I(0), variable? 10 The authors declare (p. 24) that “We then perform co-integration analysis. . . ”. The point is that apart from some theoretical issues, I did not trace such an analysis in the way it is applied here. Besides, nothing is mentioned about any ECVAR, analogous to the one that is formulated and used in this study. Most essentially, I have not traced, In case that the disequilibrium errors are computed from: then the co-integration vector will have the form: u i = Ŷ i − Y i (13) [−10.289613 0.003621]. (14) It is noted also that hat (ˆ) is omitted from the disequilibrium errors u i , for avoiding any confusion with the OLS disturbances. It should be emphasized that in such a case, i.e., using the disequilibrium errors computed from Eq. (13), then the sign of the coefficient of adjustment, as will be seen later, will be changed. 10 It should be re-called at this point, that if t i is a common trend, then t i = 1(∀ i).
214 Athanasios Athanasenas in relevant works, the stability analysis applied for the system presented in Eq. (8). 4.5. The ECM Formulation The lagged values of the disequilibrium errors, that is u i−1 , serve as an error correction mechanism in a short-run dynamic relationship, where the additional explanatory variables may appear in lagged first differences. All variables in this equation, also known as ECM, are stationary so that, from the econometric point of view, it is a standard single equation model, where all the classical tests are applicable. It should be noted, that the lag structure and the details of the ECM, should be in line with the formulation as seen in Eq. (6a). Hence, I started from this relation considering the errors u i and estimated the following model, given that the maximum lag length is p − 1 i.e., 3. Y i = α 0 + 3∑ α j Y i−j + j=1 3∑ β j W i−j − a Y u i−1 + Y v i . (15) j=1 Note that Y v i are the model disturbances. If the adjustment coefficient a Y is significant, then we may conclude that in the long run, W causes Y. If a Y = 0, then no such a causality effect exists. In case that all β j are significant, then there is a causality effect in the short run, from W to Y. Ifallβ j = 0, then no such causality effect exists. The estimation results are presented below. Y i = 0.5132+ 0.2618Y i−1 + 0.161Y i−2 − 0.00432Y i−3 + 0.0273W i−1 (0.125) (0.074) (0.073) (0.0738) (0.075) p value 0.0001 0.0005 0.0276 0.953 0.719 Hansen 0.0370 0.1380 0.2380 0.234 0.128 + 0.0487W i−2 − 0.076W i−3 − 0.0879u i−1 + Y ˆv i (0.075) (0.065) (0.022) p value 0.519 0.242 0.0001 Hansen 0.065 0.098 0.025 (for all coefficients 2.316) (16) ¯R 2 = 0.167,s= 0.009, DWd = 1.98,F (7,189) = 6.65, Condition number (CN) = 635.34. (p value = 0.0)
- Page 185 and 186: 162 Anna-Maria Mouza Thus, the tota
- Page 187 and 188: 164 Anna-Maria Mouza 6. Some Altern
- Page 189 and 190: 166 Anna-Maria Mouza increase for e
- Page 191 and 192: 168 Anna-Maria Mouza The solution o
- Page 193 and 194: This page intentionally left blank
- Page 195 and 196: This page intentionally left blank
- Page 197 and 198: 174 Fabrizio Iacone and Renzo Orsi
- Page 199 and 200: 176 Fabrizio Iacone and Renzo Orsi
- Page 201 and 202: 178 Fabrizio Iacone and Renzo Orsi
- Page 203 and 204: 180 Fabrizio Iacone and Renzo Orsi
- Page 205 and 206: 182 Fabrizio Iacone and Renzo Orsi
- Page 207 and 208: 184 Fabrizio Iacone and Renzo Orsi
- Page 209 and 210: 186 Fabrizio Iacone and Renzo Orsi
- Page 211 and 212: 188 Fabrizio Iacone and Renzo Orsi
- Page 213 and 214: 190 Fabrizio Iacone and Renzo Orsi
- Page 215 and 216: 192 Fabrizio Iacone and Renzo Orsi
- Page 217 and 218: 194 Fabrizio Iacone and Renzo Orsi
- Page 219 and 220: 196 Fabrizio Iacone and Renzo Orsi
- Page 221 and 222: 198 Fabrizio Iacone and Renzo Orsi
- Page 223 and 224: 200 Fabrizio Iacone and Renzo Orsi
- Page 225 and 226: 202 Athanasios Athanasenas 2. The C
- Page 227 and 228: 204 Athanasios Athanasenas to asset
- Page 229 and 230: 206 Athanasios Athanasenas I end up
- Page 231 and 232: 208 Athanasios Athanasenas consider
- Page 233 and 234: 210 Athanasios Athanasenas Table 1.
- Page 235: 212 Athanasios Athanasenas vector:
- Page 239 and 240: 216 Athanasios Athanasenas The esti
- Page 241 and 242: 218 Athanasios Athanasenas and ⎡
- Page 243 and 244: 220 Athanasios Athanasenas run, cre
- Page 245 and 246: 222 Athanasios Athanasenas Lown, C
- Page 247 and 248: 224 Index fiscal expansions, 91, 92
Credit and Income 213<br />
4.4. Essential Remarks<br />
In seminal works about money and income (as for instance in Tobin, 1970),<br />
the analysis is restricted in a pure theoretical consideration. Some authors<br />
(see for instance, Friedman and Kuttner, 1992) write many explanations<br />
about co-integration, co-integrating vectors, and error-correction models,<br />
but I did not trace any attempt to compute either. In a later paper by the<br />
same authors (Friedman and Kuttner, 1993), I read (p. 200) the phrase,<br />
“vector autoregression system,” which in fact is understood as a VAR. The<br />
point is that no VAR model is specified in this work. Besides, the lag length<br />
is determined by simply applying the F-test (p. 191). In this context, the<br />
validity of the results is questionable. In other applications (see for instance,<br />
Arestis and Demetriades, 1997), where a VAR model has been formulated,<br />
nothing is mentioned about the test applied to determine the maximum lag<br />
length. Also, in this paper, the authors are restricted to the use of trace<br />
statistic (l (trace) ) to test for co-integration (p. 787), although this is the<br />
necessary but not the sufficient condition, as pointed out above. From other<br />
research works (see for instance, Arestis et al., 2001), I get the impression<br />
that there is not a clear notion regarding stationarity and/or integration,<br />
since the authors state [p. 22 immediately after Eq. (1)], that “D isaset<br />
of I(0) deterministic variables such as constant, trend and dummies. . . ”.<br />
It is clear that this verification is false. How can a trend, for instance, be<br />
a stationary, I(0), variable? 10 The authors declare (p. 24) that “We then<br />
perform co-integration analysis. . . ”. The point is that apart from some<br />
theoretical issues, I did not trace such an analysis in the way it is applied<br />
here. Besides, nothing is mentioned about any ECVAR, analogous to the one<br />
that is formulated and used in this study. Most essentially, I have not traced,<br />
In case that the disequilibrium errors are computed from:<br />
then the co-integration vector will have the form:<br />
u i = Ŷ i − Y i (13)<br />
[−10.289613 0.003621]. (14)<br />
It is noted also that hat (ˆ) is omitted from the disequilibrium errors u i , for avoiding any<br />
confusion with the OLS disturbances.<br />
It should be emphasized that in such a case, i.e., using the disequilibrium errors computed<br />
from Eq. (13), then the sign of the coefficient of adjustment, as will be seen later, will be<br />
changed.<br />
10 It should be re-called at this point, that if t i is a common trend, then t i = 1(∀ i).