Instrumental variable

[Behrouz Jahan]

I am exploring the impact of social media WOM on firm performance during the 8 quarters of 2014-2015 for several firms. As WOM is likely endogenous, I was thinking of using weather data as instrument. The idea is that when, in a given quarter, the number of rainy days (or the number of days with above average precipitation) is higher, people stay at home more and are likely to, say, tweet more (relevance). Moreover, weather does not seem to directly impact firm performance (exclusion restriction).

My question is whether you think this can be a valid instrument. To be more specific, is it OK that the value of this instrument for any given quarter is identical for all firms in my sample? Thanks in advance!

[Phil Bromiley]

Welcome to Statalist. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stat output, and sample data using dataex.

While it is probably OK if the instrument is the same for multiple observations (assuming you don't want a quarter dummy), it is likely to be a weak instrument, i.e., only slightly correlated with WOM. Whether weather influences firm performance would depend on the kind of firm. xtivreg2 provides built in tests of weak instruments and other potential problems.

[Behrouz Jahan]

Thanks so much Phil. What happens if I include quarter dummies as well? Please note that I have two years of data, so 1st quarter of 2014 and 2015 will have the same quarter dummy, but different values for the temperature variable.

Welcome to Statalist. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stat output, and sample data using dataex.

While it is probably OK if the instrument is the same for multiple observations (assuming you don't want a quarter dummy), it is likely to be a weak instrument, i.e., only slightly correlated with WOM. Whether weather influences firm performance would depend on the kind of firm. xtivreg2 provides built in tests of weak instruments and other potential problems.