CARIFORUM and UK EPA Study

;

r

(1) where Yt is the log of real gross domestic product , Kt is for the log of capital (real gross domestic capital formation), Lt is for the total labor force (age 15-60) in Pakistan, Xt is for the log of aggregate real exports, πt is for the annual percentage change in the Consumer Price Index (CPI), , is for the log of domestic credit to the private sector as a percentage of GDP, and εt is for the usual stochastic error expression, assumed to be independently and identically distributed with zero mean and constant variance. It is rather apparent that the typical income-export specification typifies dual causality. We exploit this relationship to examine the effects of the covariates and income per capita on aggregate exports (the response or endogenous variable). The respecification (Equation 3) of earlier models emphasizes the importance of exports and the variables that are most likely to induce or hinder the promotion of exports. Though the respecification or relationship among the variables can theoretically be taken for granted, we plan to subject the concept of dual causality to a Granger Causality test in order to econometrically affirm that some or all of the variables may manifest interdependent effects. The Granger-Causality test, with the accompanying cointegration extension, is conventionally and pervasively utilized to establish such a relationship; see Engle Granger (1987), Rubin (2004), Ramirez (2006), Torayeh (2011), Hill et al. (2011), and Warburton (2018). The Granger test estimates pairs of regressions, say: 0 1 2 3 4 5 t t t t t t t K L X C b b b b b p b e = + + + + + + r t C Y

n

n

1 å å å å i a t i - t i = j n n i l t i - X X = + +

Y

i t j Y u b - +

=

1

t

t

;

X

i t j Y u d - +

=

2

t

t

(2)

1

1

i

j

=

=

the disturbances, u 1t and u 2t are presumed to be uncorrelated. Granger Causality

In panel workfile settings, EViews performs panel data specific causality testing. In these settings, least squares regressions can take a number of different forms, depending upon assumptions made about the structure of the panel data. Since Granger Causality is computed by running bivariate

Page 213 of 241