7.1.2Model 2: The Productivity of IT Employment at the County Level

This model explores production functions for counties in order to relate county labor productivity to some measures of IT vs. non-IT location of activity. Whereas the previous model was at the state level, this one is at the county level. I intend to estimate the relative effect of the location of IT vs. non-IT activities on county productivity using the location quotient measure of county differences in industrial mixes. Although this measure is usually used when many different industries are considered, the technique is also applicable when only two types of industries are considered, namely the IT and non-IT types of industries. The location quotient for state s (LOCs) is defined as:

LOCs,e = (nc,e / nc) / (ns,e / ns) (7.6)

Two location quotients will be computed for each county, for each type of employment, IT and non-IT. A high value of the location quotient for one type of industry in a given county indicates that activity in that industry is more intense in that county compared to the overall state intensity in that industry. This measure is conceptually related to the concentration ratio previously defined, but does not take the sizes of counties into account and is computed for each industry in each county.

My goal is to compare the relative effect of IT concentration compared to non-IT concentration on productivity. To do so, I will use the ratio of the two location quotients to evaluate the effect of one compared to the other. This ratio is expressed by

RLOC= LOC1,c / LOC2,c (7.7)

If RLOC increases in a county, then it means that IT activity in the county is more concentrated, or that non-IT activity is less concentrated. I will introduce this measure of activity into an equation estimating county labor productivity, based on a Cobb-Douglas specification:

ln pc = lnA + αc lnknc + μlnRLOCc + ΣDs-1(7.8)

where Ds-1 is a set of state dummy variables controlling for state fixed effects.

Finding a positive and significant estimate for parameter μ will indicate a positive effect of the ratio of location quotients on productivity. In other words, it would mean that if the location quotient of IT activity in a county increases relative to the location quotient of non-IT activity in that county, then labor productivity should increase in that county. If results go in the same direction as those found using model 1, then conclusions relative to the location of IT vs. non-IT activity and its effect on productivity will be more reliable. I will now turn to the third model, which is an attempt to link county activity location patterns to state productivity.