8.3IT vs. Non-IT Location Quotients and Labor Productivity (Model 2)

This section presents the results of the analysis, at the county level, of the localization economies or diseconomies arising from IT and non-IT employment. Equations estimates are based on the model described in equation 8.4, with location or density quotients in place of concentration rate. Estimates appear in Table 8.3.

Looking at Table 8.3, it appears that all coefficients are significant at the 0.01 level, except for regression (8.9), where multicollinearity between lnloc1,c and lnloc2,c may corrupt the results. This is confirmed by a strong positive coefficient of correlation of –0.86 between these two variables. Therefore, independent regressions for each of those have to be considered. Regression (8.7) indicates a positive coefficient of 0.071 for the IT location quotient variable. On the other hand, the coefficient for the non-IT location quotient is negative in regression (8.8), with a value of –0.107. This means that if the county percentage of IT employee doubles, holding the state percentage constant, county labor productivity should increase by 7.1%. In other words, if a county “imports” an extra 100% of IT employees from other counties in the same state, labor productivity in that county should increase by 7.1%. On the other hand, if a county does the same regarding its non-IT employment level, its labor productivity may decrease by more than 10%. These results can be interpreted once again as reflecting externality effects, reaching conclusions similar to those in the previous section about employment concentration. Indeed, IT employment may be subject to agglomeration effects stronger than congestion effects, and the reverse should be true regarding traditional employment.

Finally, the ratio of the two location quotients is used as an explanatory variable in regression (8.10). The resulting coefficient is positive and strongly significant, with a value of 0.046. This ratio could be simplified as shown in the following equation:

rloc = loc1,c / loc2,c = {[(n1,c/nc)/(n1,s/ns)] / [(n2,c/nc)] / (n2,s/ns)]}

⇔ rloc = n1,c . nc -1. n1,s -1. ns . n2,c -1 . nc .n2,s . ns -1

⇔rloc = (n1,c / n1,s ) / (n2,c / n2,s)(8.5)

Table 8.3Estimates of Elasticities from Equations of Models 2 and 3: Location and Density Quotients and Productivity
Regression # (8.6) (8.7) (8.8) (8.9) (8.10) (8.11) (8.12)
Constant 6.034
(72.27)		 6.142
(75.30)		 6.141
(75.83)		 6.142
(75.79)		 6.143
(75.64)		 6.163
(71.21)		 6.047
(69.75)
Capital-to-output ratio 0.370
(82.14)		 0.381
(84.57)		 0.382
(85.72)		 0.382
(85.38)		 0.382
(85.31)		 0.372
(81.68)		 0.369
(81.59)
Education 0.152
(9.45)		 0.102
(6.33)		 0.097
(6.10)		 0.097
(6.07)		 0.097
(6.05)		 0.114
(6.39)		 0.150
(8.53)
IT location quotient 0.071
(11.88)		 0.002
(0.23)
Non-IT location quotient -0.107
(-13.58)		 -0.104
(-6.60)
IT density quotient 0.005
(4.26)
Non-IT density quotient -0.000
(-0.01)
Ratio of location quotients 0.046
(13.01)
R2-adjusted 0.79 0.80 0.80 0.80 0.80 0.79 0.79
F-stat 231 240 245 240 244 229 228
Note: All variables are significant at the 0.01 level except for the coefficient on IT location quotient in regression (8.9). All regressions include (1-s) state dummy variables, which are almost all significant at the 0.01 level

Thus the ratio of location coefficients may be interpreted as the percentage of IT employment in a county relative to the state percentage, divided by the same percentage measure for non-IT employment. Then, the value of the coefficient for the ratio of location quotients expresses the percentage of change in county labor productivity due to an increase in the county percentage of state employment in IT relative to the county percentage of state employment in traditional industry. Hence, given the value of 0.046, it must be that if IT employment doubles in a county, holding traditional employment constant, 4.6% gains in labor productivity would be observed in this county as a result.