THE LONG RUN RELATIONSHIP BETWEEN KNOWLEDGE AND WAGES IN THE MANUFACTURING SECTOR IN PUERTO RICO

Diego IRIBARREN (CRIC & Estudios Técnicos, Inc.)

Following up on the hypothesis that knowledge must remain invariantly stable in order for dependency between any two series to be detectable, we study the cointegration of the average salary and knowledge in the manufacturing sector in Puerto Rico. The aim was to determine whether changes in the average salary were correlated in the long run with changes in knowledge. To do this, we first estimated a model for the level of employment in the sector using the Kalman filter algorithm. The time series of knowledge was estimated as the state variable of the entire system. Two general specifications were considered, one with stochastic regressors and one with completely deterministic ones. Several covariance structures were considered as well to determine the best model. In this manner, several knowledge time series were constructed. We then studied whether the growth rate of each one of the knowledge time series was cointegrated with the growth rate of the average salary in the sector. Some non-standard tests, i.e. non-parametric, were used in both the unit root and the cointegration phase in order to generalize the specifications. The results seemed to indicate that changes in the average salary do indeed hold a long-term stationary relationship with knowledge thus providing the required structure from the initial hypothesis.