THE
CENTER FOR THE STUDY OF INDUSTRIAL ORGANIZATION
at
NORTHWESTERN
UNIVERSITY
         


Home



People



Working Papers



Seminars and Conferences



Data Sets and Computer Code



Links



E-mail List and Contact

Adoption of 56K Modems by ISPs Dataset


This data set contains information on the adoption of 56K modems by Internet Service Providers in October of 1997. For each ISP, we observe which of two modem technologies the ISP adopted. The data set indicates every telephone switch that can call each ISP so we have a very good measure of the number of competitors faced by each ISP. ISPs vary by size and other technologies, both of which are observed in the data. In addition, there is local demographic data associated with each switch.

domain.dta

domaineea.dta

eea.dta

estimationdata.dta

This data set was used in the following papers:

Augereau, A., Greenstein, S., & Rysman, M. (in press). Coordination vs. Differentiation in a Standards War: 56K Modems. RAND Journal of Economics.

Rysman, M., & Greenstein, S. (2005). Testing for Agglomeration and Dispersion. Economics Letters, 86 , 405-411.

Greenstein, S. & Rysman, M. (in press). Coordination Costs and Standard Setting: Lessons from 56K Modems. In S. Greenstein & V. Stango (Eds.), Standards and Public Policy. Cambridge University Press. [Also CSIO Working Paper #0056, Northwestern University.]

Goldfarb, Avi, and Botao Yang. Are All Managers Created Equal? Working Paper. University of Toronto.

Construction of the data:

From theDirectory, we obtain a list of every telephone number that can be used to access each Internet Service Provider. From Boardwatch, we learn the 56K adoption choice of each ISP. We match these data sets together to obtain our final list of ISPs. From CCMI data, we match telephone numbers to telephone switches and we determine which telephone switches can call which for free under the local calling plan. When there is a choice, we use the lowest cost local calling plan available from the incumbent telephone company. In this way, we determine every switch that can call each ISP for no charge.

Much of this data is captured in the file DomainEEA. The file contains one observation for every ISP-switch.

EEA is a telephone switch identifier. It normally corresponds to a single instance of the first 6 numbers of a 10 digit telephone number. For instance, 617-353 might be one switch. This is not exact because consumers move and carry their phone numbers with them.

FIPSCODE is the Census identifier of the county that the switch resides in.

DOMAIN is the name of the ISP, which we use as the ISP identifier. An EEA-DOMAIN combination determines a unique observation in the data. Naturally, a given DOMAIN may show up multiple times if multiple switches can call it (typically the case even if a DOMAIN has only one telephone number) and a given EEA can show up multiple times if a switch is served by multiple ISPs.

PREVCH (previous choice) is an indicator for the adoption choice of the ISP in July 1997. We use the July issue of Boardwatch to fill out this variable for ISPs that we observe in October. That is, we have no observations for ISPs that appear in July and not in October.

N_ISP (number of ISP's) is the total number of ISP appearing in the switch in the theDirectory data set. Boardwatch contains data on only about half of the ISPs in theDirectory so this number is higher than if you simply used the file DomainEEA to construct this count.

SUMBACK is the total number of Internet backbone providers at the switch.

ADOPT is the adoption choice of the ISP in October 1997. For both ADOPT and PREVCH, 0 means no adoption, 1 means the ISP adopted FLEX, 2 means the ISP adopted X2 and 3 means the ISP adopted both.

The file DOMAIN contains further ISP information. There is one observation for every ISP.

DOMAIN indicates the ISP name.

BANDWDTH indicates the speed at which the ISP can download data from the Internet.

PORTS indicates the number of ports that an ISP has, which is the total number of consumers the ISP could service at a time.

T1 is a dummy for whether the ISP has a T1 line to connect to the Internet. A T1 line provides a very fast and broad internet connection and would signal a large or high quality ISP.

ISDN is a dummy for whether an ISP offers ISDN service to consumers. An ISDN line is also a fast connection to the Internet, although not as fast as a T1 line. An ISP must have an ISDN connection to the Internet to offer ISDN service to consumers.

We felt that PORTS was poorly constructed because it did not always correspond to our notion of size. For instance, compare PORTS to the number of switches an ISP is in.

The file EEA contains further switch level information. There is one observation for every switch (EEA).

EEA is a telephone switch identifier.

ZIPCODE indicates the 5 digit zip code that the switch is in. We could not match to the 5 digit zip code in a few cases and instead matched to the 3 digit zip code.

The rest of the variables have descriptive labels and are demographic data extracted from the census.