Labour Economics 11 (2004) 119 – 127 www.elsevier.com/locate/econbase
Department of Economics, University of Hawaii at Manoa, 2424 Maile Way, Saunders 542,
Honolulu, HI 96822, USA
Abstract
By using recent Current Population Survey (CPS) data sets, this paper investigates whether workers who use the Internet at work earn a higher wage than otherwise similar workers who do not use the Internet at work. Estimates suggest that there was a substantial premium to Internet use in
1997 of about 8%. However, this premium disappeared very quickly thereafter, and there are grounds to believe that it is now negative. This result is in stark contrast with that for computer use.
D
2003 Elsevier B.V. All rights reserved.
JEL classification: J31; O33
Keywords: Internet; Computer; Wage structure
1. Introduction
The rapid development and diffusion of new information technologies such as the
Internet and computers has altered the production process in many workplaces. It is also widely believed that the introduction of these new technologies has altered the wage structure in the labor market. For example, in his influential study,
Krueger (1993)
finds that workers who use computers on their job earn 10 – 15% higher wages than non-users.
Additionally, he found that the expansion of computer use in the 1980s accounts for onethird to one-half of the increase in the rate of return to education.
An interesting question might be whether, if any, a premium for Internet use is different from that for computer use. It is theoretically ambiguous whether Internet use can account for the changes in the wages of various types of workers. However, at least two characteristics of Internet use are noteworthy. On one hand, the Internet is a white-collar office tool that requires less skill to use than might be required for any other kind of
* Corresponding author. Tel.: +1-808-956-8590; fax: +1-808-956-4347.
E-mail address: leesang@hawaii.edu (S.-H. Lee).
0927-5371/$ - see front matter
D
2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0927-5371(03)00052-6

120 S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 computer use. A ballet dancer could send e-mails to her friends or finds information at the website with little knowledge about it, which would not affect her earnings as a ballet dancer. On the other hand, the vast use of Internet since mid-1990s provided ample opportunities for business. The Internet has revolutionized the communication and computer world like nothing before.
1
While we do not have any priors regarding the degree to which Internet specific technological change is linked to changes in wage structure, we may not totally ignore the possibility of a rent-sharing mechanism, which would affect the wage structure.
In this paper, we estimate the premium to Internet use using recent Current Population
Survey (CPS) data sets. Since these data sets include information on both Internet use and computer use, we are able to compare the effects of the use of these two prominent information technologies on the wage structure. The ordinary least squares (OLS) estimates suggest that there was a substantial premium to Internet use in 1997 of about
8%, and it is significant. However, this premium disappeared very quickly thereafter, and there are grounds to believe that it is now negative. This result is in stark contrast with that for computer use, because our results also suggest that the premium to computer use found in Krueger has been persistent in the 1990s. The remainder of the paper is organized as follows. The next section describes the data used in this study. Section 3 presents our empirical results. The final section discusses the interpretation of these results.
2. Data and descriptive analysis
The analysis in this study relies primarily on data sets from the CPS conducted in
October 1997, December 1998, August 2000, and September 2001. The information regarding computer use at work is based on October 1997 and September 2001 surveys, whereas the information of Internet use at work is based on surveys taken in October 1997,
December 1998, and August 2000. Only the October 1997 CPS has information on both computer use at work and Internet use at work. In order to compare the results with that of
Krueger (1993) , the data set is manipulated in the same way in Krueger.
2
Table 1
reports the proportion of workers who use the Internet and computers at work.
We report two measures of proportion. One is the percentage of workers using the Internet at work, which we call the unconditional use of the Internet. The other is the percentage of
1
The possible benefits from an Internet-enabled transformation of business organization are enormous. For example, the Internet provides ample scope for cost reduction across all stages of the production process. The
Internet can also drastically accelerate speed-to-market by reducing time it takes to transmit. The low cost of expanding a functioning network is a further advantage. For example, by moving customer service and technical support online, the Cisco Systems, a leading networking equipment vendor, increased productivity by 200 –
300%, resulting in savings of US$125 million in customer costs.
2
The sample is restricted to individuals between ages 16 and 65. The hourly wage is the ratio of usual weekly earnings to usual weekly hours. Individuals who earned less than US$1.50 per hour or more than US$200 per hour are deleted from the sample. The weekly earnings variable in the 1997 CPS is top-coded at US$1923, whereas the weekly earnings in the other surveys are top-coded at US$2884.01. In order to circumvent any problems caused by changes in top-coding over time, we calculated an estimate of the mean log hourly wage for individuals who were top coded in 1998 and assigned it to each individual who was top-coded in 1997. See
Appendix A in Krueger for details.

S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 121
Table 1
Percent of workers in various categories who use internet at work/computer at work
Use the internet at work Use a computer at work
(1)
Oct.
1997
(2)
Dec.
1998
(3)
Aug.
2000
(4)
Oct.
1984
(5)
Oct.
1989
(6)
Oct.
1993
(7)
Oct.
1997
(8)
Sep.
2001
Internet at work/ computer at work
(9) (10) (11)
1997/
1997
1998/
1997
2000/
2001
All workers
Gender
Men
Women
16.2
17.1
15.2
Education
< High school 0.7
High school 5.8
Some college 15.7
z
College
Race
35.4
White
Non-white
Age
Age 18 – 24
16.9
10.1
Age 25 – 39
Age 40 – 54
Age 55 – 64
Occupation
9.1
17.9
18.7
13.2
24.5
24.4
24.7
4.9
13.1
25.8
43.6
25.2
18.6
29.0
26.6
22.5
17.0
29.1
27.7
30.6
4.2
16.7
29.9
50.7
29.8
22.7
31.5
31.3
28.4
20.5
25.1
21.6
29.6
5.1
19.2
30.6
42.1
25.8
18.6
20.5
29.6
23.9
17.7
37.4
32.2
43.8
7.7
28.4
45.0
58.5
38.5
28.1
29.6
41.4
38.9
27.0
46.6
41.1
53.2
10.4
34.6
53.1
70.2
48.0
36.7
34.3
49.8
50.0
37.3
48.2
42.8
54.2
6.4
31.8
54.2
74.3
49.6
36.6
34.0
51.6
53.3
43.1
52.5
47.2
58.4
8.2
34.5
56.7
78.3
54.0
40.6
34.8
55.9
57.7
51.6
33.6
40.0
28.0
10.9
18.2
29.0
47.6
34.1
27.6
26.8
34.7
35.1
30.6
50.8
57.0
45.6
76.6
41.2
47.6
58.7
50.8
50.8
85.3
51.6
42.2
39.4
55.4
58.7
52.4
51.2
48.4
52.7
64.8
55.2
55.9
90.5
56.0
49.2
39.7
Blue-collar 3.6
White-collar 25.4
Union member
Yes
No
Hours
12.9
16.3
Part-time
Full-time
Region
Northeast
7.5
18.1
Midwest
South
West
Number of observations
9.9
33.9
21.7
24.6
22.8
24.9
12.7
40.0
25.0
29.2
22.4
30.2
7.1
39.7
19.9
25.3
14.8
29.3
11.2
56.6
31.8
37.7
24.4
42.3
17.1
67.6
39.1
46.9
29.3
51.0
20.1
69.7
45.6
48.3
31.7
51.7
24.0
72.8
51.0
52.5
38.2
55.4
17.9
36.4
28.3
33.7
23.7
35.0
49.3
48.6
47.6
50.9
71.9
48.2
52.9
54.9
49.0
55.6
58.6
54.5
15.8
15.4
15.5
18.1
23.3
25.3
23.1
26.5
28.4
30.1
27.4
30.5
25.5
24.3
23.2
28.9
37.6
36.6
36.6
39.7
46.9
46.7
45.0
48.8
48.4
48.8
46.9
48.9
53.3
53.3
51.1
52.6
32.6
31.6
33.0
37.0
60,091 60,196 60,698 61,704 62,748 59,852 60,091 71,090 –
48.1
51.8
49.3
54.2
–
53.3
56.5
53.6
58.0
–
Columns (4), (5), and (6) are from the Table 4 of Autor et al. (1998) . The other columns are the authors’
tabulations. Data for all columns are from the Current Population Survey.
computer users who use the Internet at work. In order to distinguish this from unconditional use of Internet, we name it the conditional use of the Internet.
Columns (1) through (8) in Table 1
summarize the unconditional use of the Internet. We also tabulate the proportion of workers who use computers at work for comparison.
Between 1997 and 2000, the percentage of workers using the Internet at work increased by about 80%, from 16% to 29%. The increase in Internet use at work during October 1997 –
December 1998 period is remarkable. We find that white and highly educated workers are

122 S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 more likely to use the Internet at work than non-white and less educated workers. Full-time and non-union members are also more likely to use the Internet. However, unlike the case for computer use, there is no clear evidence that women are more likely to use the Internet at work. The most noticeable differences between Internet use and computer use occur when one segregates workers by age. While workers aged 40 – 54 are more likely to use computers at work than those aged 18 – 25, younger workers were more likely to use the
Internet than older workers since 1998.
The difference between Internet use and computer use can be seen more clearly by the conditional Internet use, which is reported in columns (9), (10), and (11). These three columns present the percentage of computer users who use the Internet. Column (9) is based on information from October 1997 CPS, which contains information on both
Internet use and computer use at work. Columns (10) and (11) are based on information of
Internet use from December 1998 and August 2000 surveys and computer use in October
1997 and September 2001 surveys. Obviously, column (10) overestimates the true value in
1998 and column (11) underestimates the true value in 2000. This is because column (10) expresses Internet use in 1998 as a share of computer use in 1997 and column (11) expresses Internet use in 2000 as a share of computer use in 2001.
The percentage of computer users who uses the Internet at work increased from 34% in 1997 to 51% in 1998, and to about 55% in 2000. These three columns show that there is a significant transition in the use of the Internet by groups over time. For example, although women were more likely to use computers than men, men were much more likely to use the Internet at work than women conditional on their use of computers at work. In 1997, only 28% of women used the Internet at work, suggesting that 72% of women used computers but did not use the Internet at work. However, the difference in Internet use between men and women has decreased drastically since then.
In 2000, there is also almost no difference between white and non-white workers in the use of Internet conditional on their use of computers. There is also no sizable difference between blue-collar and white-collar workers in using the Internet in 2000, although white-collar workers were more likely to use the Internet at work in 1997. Even part-time workers are not less likely to use the Internet than full-time workers in 2000. The only exception is the Internet use by age group; younger workers were more likely to use the
Internet than older workers conditional on their use of computer at work. Thus, these results suggest that the between group variation in the use of Internet in columns (2) and
(3) is mostly due to the between group variation in the use of computers, with one exception of its correlation with age. Thus, the results suggest that Internet use inequality mostly disappeared by 2000, while the computer use inequality remains persistent.
3. Internet use and wages
We use the following two statistical models to estimate the premiums to Internet use.
ln w ¼ x b þ c I þ e ln w ¼ x b þ c I þ y
C þ e
ð
ð
1
2
Þ
Þ

S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 123 where ln w represents log wage, x represents the worker’s observed characteristics, I(C) is a dummy variable that equals one if the worker uses the Internet (computers) at work. In
Eq. (1), the base category of the Internet dummy includes two types of workers, those who do not use computers at work and those who use computers at work without using Internet.
We name it the unconditional wage premium to Internet use. Because the 1997 CPS contains information on both Internet use at work and computer use at work, we can also estimate Eq. (2), the coefficients of both Internet use and computer use. In this case, the coefficient on Internet use should be interpreted as indicating the additional payoff associated with Internet use relative to any computer use. In order to distinguish this from unconditional premium, we name the additional payoff the conditional premium.
Both equations are estimated using the OLS method.
Columns (1) through (3) of Table 2
report results of fitting Eq. (1). The results suggest that workers who use the Internet while on the job receive higher wages than those who do not by 13% (exp(0.126) 1) in 1997, 9% in 1998, and 5% in 2000. These results are all statistically significant at the 1% significance level. When we include 45 two-digit occupation dummy variables, the estimated coefficients are 0.114 in 1997, 0.077 in
1998, and 0.045 in 2000 and they are highly significant.
Column (4) presents our results on conditional premium to Internet use. The results indicate that workers who use the Internet at work receives additional wages that are 8% higher (exp(0.074) 1) than those who do not use Internet at work. Krueger shows that in
1989 the most highly rewarded type of computer use was electronic mail. Since Krueger’s comparable estimate for the same specification was 7% in 1989, we conclude that the reward for using the Internet did not decrease at all until 1997.
We also estimate the unconditional premium to computer use by using 1997 and 2001 data sets. That is, we re-estimate Eq. (1) replacing the Internet dummy with a computer dummy. The results are reported in columns (5) and (6) of the table. The earnings differential in hourly pay between workers who use computers on the job and those who do not is 14% in 1997 and 11% in 2001, suggesting that there is still substantial
(unconditional) earnings premium to computer use. Because Krueger’s same measures
3 are 15% in 1984 and 18% in 1989, we conclude that there is only a slight decrease in premiums to a computer use between 1984 and 2001. This is in stark contrast with the very rapid decrease in the premium for Internet use.
Although only the 1997 CPS contains information on both Internet use at work and computer use at work, we can still calculate the approximate conditional premium to
Internet use in 1998 and 2000 using the proportion of Internet users and the information on unconditional premium provided in
Tables 1 and 2 , respectively.
Again, the unconditional premium is the premium against two types of workers, the premium against those who do not use computers at work and the premium against
3
Two covariates of our specification are different from Krueger. First, instead of years of schooling, we use four education category variables due to a change in variable definition since 1994. Second, as a result, the potential experience variable in Krueger is replaced with an age variable. To determine whether this change in specification affects the overall result, we used the October 1989 survey and regress the wage equation first with using years of schooling and potential experience, and then with using the education category variables and age.
The results are almost identical, suggesting that using education category variables instead of years of schooling barely affect our estimated coefficients.

124 S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127
Table 2
OLS regression estimates of the effect of internet and computer use on pay
(1)
Oct. 1997
(2)
Dec. 1998
(3)
Aug. 2000
(4) (5)
Oct. 1997 Oct. 1997
(6)
Sep. 2001
Use internet at work
Use computer at work
0.126
(0.015)
–
High school
Some college
College
Graduate
Age
Age
2 H 100
Black (1 = yes)
Other race (1 = yes)
Par-time (1 = yes)
Lives in SMSA (1 = yes) 0.056
(0.008)
Veteran (1 = yes) 0.027
(0.015)
Female (1 = yes)
Married (1 = yes)
0.080
(0.013)
0.114
(0.012)
Married*female
Union member (1 = yes)
R
2
Number of observation
0.097
(0.016)
0.247
(0.012)
0.458
7766
0.186
(0.018)
0.249
(0.019)
0.367
(0.022)
0.465
(0.035)
0.039
(0.002)
0.042
(0.003)
0.078
(0.013)
0.012
(0.019)
0.125
(0.011)
0.083
(0.011)
–
0.196
(0.017)
0.257
(0.018)
0.396
(0.022)
0.504
(0.033)
0.034
(0.002)
0.035
(0.003)
0.046
(0.013)
0.031
(0.018)
0.171
(0.011)
0.060
(0.008)
0.020
(0.015)
0.113
(0.012)
0.095
(0.013)
0.046
(0.016)
0.212
(0.012)
0.477
7739
0.052
(0.010)
–
0.158
(0.017)
0.217
(0.018)
0.333
(0.022)
0.439
(0.032)
0.039
(0.002)
0.042
(0.003)
0.026
(0.014)
0.048
(0.018)
0.134
(0.011)
0.075
(0.008)
0.057
(0.016)
0.077
(0.012)
0.099
(0.012)
0.076
(0.016)
0.188
(0.012)
0.459
7966
0.074
(0.016)
0.113
(0.010)
0.168
(0.018)
0.218
(0.019)
0.333
(0.022)
0.431
(0.034)
0.038
(0.002)
0.041
(0.003)
0.067
(0.013)
0.020
(0.019)
0.111
(0.011)
0.056
(0.008)
0.026
(0.015)
0.087
(0.012)
0.113
(0.012)
0.099
(0.016)
0.234
(0.011)
0.466
7766
– –
The dependent variable is log hourly wage. Other independent variables include eight occupation dummies.
Standard errors are shown in parentheses.
0.019
(0.019)
0.113
(0.011)
0.057
(0.008)
0.027
(0.015)
0.090
(0.012)
0.111
(0.012)
0.099
(0.016)
0.248
(0.012)
0.465
7766
0.129
(0.010)
0.171
(0.018)
0.222
(0.019)
0.338
(0.022)
0.440
(0.034)
0.038
(0.002)
0.041
(0.003)
0.068
(0.013)
0.014
(0.017)
0.142
(0.010)
0.059
(0.008)
0.046
(0.014)
0.087
(0.012)
0.084
(0.012)
0.082
(0.015)
0.198
(0.011)
0.437
9171
0.102
(0.009)
0.183
(0.018)
0.222
(0.019)
0.372
(0.022)
0.510
(0.032)
0.033
(0.002)
0.034
(0.003)
0.045
(0.013) those who use computers at work without using the Internet. Therefore, the following relationship between the unconditional and conditional wage premiums holds, given a condition that there is little interaction between the dummy indicators and the other variables.
c
I u
¼
1 P ð C Þ
1 P ð I Þ
ð c
I = C c
þ c
C c
Þ þ
P ð C Þ P ð I Þ
1 P ð I Þ c
I = C c
ð 3 Þ

S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 125 or c I = C c
¼ c I u
1 P ð C Þ
1 P ð I Þ c C c
ð 3 V where c
I u is the unconditional Internet premium, c
I/C c conditional Internet premium, c
C c premium to any (general) computer use, P (C) proportion of workers using computers at work, and P (I) is the proportion of workers using the Internet at work. Furthermore, if a fraction of computer users use the Internet, then the true c
C c is given by
4 c C c
¼ c C u q c I = C c
ð 4 Þ where c
C u is the unconditional computer premium and q is the proportion of computer users who use the Internet (thus q = P (I)/ P (C)). Plugging Eq. (4) into Eq. (3 V ) yields c
I = C c
¼
1 qP ð C Þ
1 q c
I u
1 P ð C Þ
1 q c
C u
ð 5 Þ
If everybody uses a computer ( P (C) = 1), then c
I u is equal to also use the Internet at work ( P (C) = P (I), q = 1), then (1 q ) c
C
.
c
I/C c
C c decreases, c c
I u
I u c
= 0.083, c
= 0.052, c u
C u
C u increases, and q increases with a condition of c u rapid decrease in c
I u
, little change in c
= 0.12,
= 0.10, q q
= 0.5, and
= 0.6, and
P
P
(C) = 0.5, we get
(C) = 0.5, we find c
I/C c c
I/C c
. If computer users equals zero, and c
. In 1998, with equal to zero. In 2000, with equal to negative 3%. The
I u is equal to c u
Eq. (5) clearly shows that the conditional Internet premium (
C I
< c
C u
.
c
5
I/C c
) decreases as
C u
, and an increase in q with c
I u
< c
C u c
Our results show a
I u calculation implies that the estimated conditional premium to Internet use decreased very quickly after 1997, and there are grounds to believe that it is now negative. This is in contrast to the slow decrease in the returns
6 to computer use during the much longer period of time.
These results might be closely related with the expansion of Internet technology.
Krueger argues that the higher premium to e-mail use in 1989 probably reflects the fact that e-mail was first introduced to higher paying or high-ranking jobs. In fact, less than 6% of workers used e-mail at work in 1989.
7
We have also observed a high degree of inequality between workers in using the Internet at work until 1997, but the inequality by groups has disappeared quickly thereafter. This strongly implies that a substantial portion of the 1997 Internet premium might be due in part to the fact that Internet use was not universal in 1997 and was instead often limited to selected workers, high-ranking or high paying workers. Krueger also argues that, given the substantial expansion in the supply of workers who have computer skills between 1984 and 1989, the absence of a decline in the
4
5
We are grateful to Gilles Saint-Paul for suggesting Eq. (4).
Holding c
I u and c
C u constant,
B c
I/C c
/
B q = ((1 P (C))/(1 q )
2
)( c
I u c
C u
), which is negative as long as c
I u
< c
C u
.
6
From Eq. (4), it is easy to see that the conditional premium to computer use is not less than unconditional premium after 1997 since c
I/C c
7 equals zero in 1998 and becomes negative thereafter.
Although the Internet and e-mail protocols were born in 1960s, it was not until 1989 that the first relay was made between a commercial e-mail carrier and the Internet. The World-Wide Web was first released in 1991
(source: Hobbe’s Internet timeline v5.6
http://www.zakon.org/robert/internet/timeline ).

126 S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 wage differential for computer use suggests that the demand for workers with computer skills may have shifted out as fast as the outward shift in the supply of computer-literate workers. Since possession of Internet skills is not as scarce as those for specific computer skills, the rapid expansion of Internet technology might have removed the premiums attached to it.
4. Discussion
One of the interesting questions raised by Krueger is whether the return to computer use observed in 1980s would persist in 1990s. Our results imply that the return to computer use decreased slowly during the 1990s and there is still substantial amount of return to computer use even as recently as 2001. However, our results from Internet use are quite different; within a very short period of time, there was nearly 80% increase in Internet use at work, and it did remove the premiums attached to it.
Neither Krueger nor this paper controls for individual unobserved heterogeneity in any direct way.
8
Several studies criticized Krueger’s work for the reason and tried control for the unobserved heterogeneity.
9
However, it is far from clear whether using a fixed-effects approach would yield a true return to the adoption of new technology. As
DiNardo and
Pischke (1997) argue, if returns to unobserved skills quickly changes in response to the introduction of a revolutionary technology, then using fixed-effects would be of little use.
On the other hand,
DiNardo and Pischke (1997)
cast doubt on the literal interpretation of the computer use wage differential as reflecting true returns to computer use. According to them, even the return to pencil use is 13%, suggesting that there is substantial selection in the use of any office tools. DiNardo and Pischke’s findings do cast some doubt on the literal interpretation of the computer use wage differential as reflecting true returns to computer use, but it does not prove there is no link between change in wage structure and computer use. Although we still do not have any clues regarding whether Krueger’s estimates represent a return to skills or a selection effect, our results do suggest that computer use is different from Internet use in terms of its effect on the wage structure. If selection is important for computers, we should expect it to be equally important for the
Internet. If returns to computer use are mostly a reflection of unobserved worker heterogeneity, it is difficult to explain the result for the Internet.
The natural question is why the return to Internet use is different from that for computer use? Although our paper is far from a clear answer, the result from the paper is somewhat consistent with an argument; the answer to the question is closely related with whether the adoption of new technology is related with skill upgrading. For example, in their plantlevel study,
Doms et al. (1997)
argue that the adoption of factory automation technologies is less correlated with skill upgrading and wages than investment in new computer
8
See the work of
Lee and Kim (2002) , which adopts Krueger’s method to control for individual unobserved
heterogeneity in the analysis of the premium to the Internet use.
9
Using a unique matched panel data set,
Entorf et al. (1999)
show that the new technologies users were already better paid before working in these jobs. Their estimates suggest that the wage increase for computer users never exceeds 2%, which is far below the cross-section estimates.

S.-H. Lee, J. Kim / Labour Economics 11 (2004) 119–127 127 equipment. Because the Internet might require less skill to use than might be required for any other kind of computer use, this may be one possible reason for our findings.
Acknowledgements
We are grateful to Jo¨rn-Steffen Pischke, Gilles Saint-Paul, and seminar participants at the European Summer Symposium in Labor Economics, the University of Hawaii at
Manoa, and the Korea Labor Institute for useful comments and suggestions. We acknowledge the use of the Internet in downloading data sets. The data set used in the paper is readily available to any researcher for the purpose of replication.
References
Autor, D.H., Katz, L.R., Krueger, A.B., 1998. Computing inequality: have computers changed the labor market?
Quarterly Journal of Economics 113, 1169 – 1213.
DiNardo, J.E., Pischke, J.-S., 1997. The returns to computer use revisited: have pencils changed the wage structure too? Quarterly Journal of Economics 112, 291 – 303.
Doms, M., Dunne, T., Troske, K.R., 1997. Workers, wages, and technology. Quarterly Journal of Economics 112,
253 – 290.
Entorf, H., Gollac, M., Kramarz, F., 1999. Technologies, wages, and worker selection. Journal of Labor Economics 17, 464 – 491.
Krueger, A.B., 1993. How computers have changed the wage structure: evidence from microdata, 1984 – 1989.
Quarterly Journal of Economics 108, 33 – 60.
Lee, S.-H., Kim, J., 2002. Bigbrother@the.office.com: Computers and the Internet, workers and wages. Working
Paper No. 02-18. Department of Economics, University of Hawaii at Manoa.