Martin Burda and Matthew Harding, "Panel Probit with Flexible
Correlated Effects: Quantifying Technology Spillovers in the Presence
of Latent Heterogeneity", Journal of Applied Econometrics,
Vol. 28, No. 6, 2013, pp. 956-981.

 
The paper uses data from

N. Bloom, M. Schankerman, and J. Van Reenen (2010): "Identifying
Technology Spillovers and Product Market Rivalry," NBER Working Paper
13060 (revise and resubmit at Econometrica) which is available at

  http://www.stanford.edu/~nbloom/replicate.zip

zipped in the file replicate.zip, containing a number of Stata files
with their detailed description. Our paper uses the specific data file
spillovers.dta, from which we extract the following variables:

    (variable name)       (description)
 1. year                  year
 2. lgspilltec1           lag log stock of tec weighted R&D (Jaffe distance) 
 3. lgspillsic1           lag log stock of sic weighted R&D (Jaffe distance)
 4. lgmalspilltec1        lag log stock of tec weighted R&D (Mahalanobis
                          distance)
 5. lgmalspillsic1        lag log stock of sic weighted R&D (Mahalanobis
                          distance)
 6. lgrd1                 lag log stock of R&D expenditures
 7. lsales1               lag log sales
 8. lgrd1_dum             missing indicator for lag log of stock R&D
                          expenditures

We further created two new variables:

 9. pat_any               binary indicator of one or more patents registered
                          in a given year (=1 if pat_count>0, =0 otherwise)
 10. firm                 unique identifier of a given firm

Thus, the dataset contains 10 variables. There are 12,928 observations
for each variable.

The dataset is provided in the file bh-data.zip, which contains the
ASCII text file rd-data.txt in DOS format. Unix/Linux users should use
"unzip -a".