How to interpret the DID result (beginner)

[Rosadalima Justina]

Hello!

I have run the DID analysis however I don't understand how to read the result of this. So, I would like to find out about the effect of mount eruption to the tourism in city 5 as the treatment compared with 4 other cities (city 6-9) as the control. I have the data from 2007-2014 where at 2010 (end) the mountain was erupted.

I have made : gen time = (year>=2011) & !missing(year) and gen treated = (city>5) & !missing(city)

Hereby I attached the result. I don't know how to interpret this and your help would means a lot. Thank you

[Rich Goldstein]

please read the FAQ to see how to post; many of us will not open binary files from strangers and some of us, including me, do not have Microsoft products

[Clyde Schechter]

I'd like to help. But please read the FAQ, especially section 12, for advice on how to show output. Microsoft Office document attachments are definitely not the way to go. Some of us don't use Office at all. But some who do, including me, will not download an Office document from a stranger because they can contain active malware. Please repost, showing your Stata results, surrounded by code delimiters so they will align nicely and be easy to read.

One comment in advance of that.

gen treated = (city>5) & !missing(city)

will distinguish city 5 from cities 6 through 9. But are there also cities 1, 2, 3, and 4? This coding will not distinguish them from city 5. Is there more to it?

Added: Crossed with Rich's response.

[Rosadalima Justina]

Oh okay, sorry my mistake. So this is the pdf file.

And to answer Clyde's question, there is no cities 1-4. So I only have cities 5,6,7,8,9.

[Clyde Schechter]

Well, you were asked to show the results (as text) within code delimiters. A PDF file is a safe download, so I took it. But others following along can't see your results and won't understand what we're talking about here. In the future please post results by copying from your log file or results window and pasting here between code delimiters, as described in section 12 of the FAQ.

The key here is the did coefficient, which is the difference in differences estimator of the "treatment" effect. It came out as 404,432.6, but with a confidence interval of -920826.9 to 1729692. So your data do not provide only a very imprecise estimate of this effect and don't even do very much to pin down the direction of the effect, let alone its magnitude. I don't think you can say much more about the results than that. Since it appears you have only one city that experienced the "treatment" and only 4 years of data on that city, it isn't surprising that your results are inconclusive. You need to find more cities that experienced that treatment that year or (since the treatment appears to have been a localized act of God) get data over a longer period of time. Or perhaps there is some different outcome variable that is less noisy and could be used instead of jumlahturis and still serve your overall purposes.

The table at the bottom of your output simply lays out the modeled values of jumlahturis in the treatment and control groups during the baseline and follow-up periods, as well as the differences between the groups in each period.

[Rosadalima Justina]

Clyde : oh okay, thank you very much!