Difference-in-Differences (DID) with varying treatment intensity at different times

[Raffaele Di Tria]

Dear all,

I am struggling to find the best way, if any, to use difference-in-differences (DID) to analyze this data.

The observations regard districts in a city, counting how many times public urban renovation has happened in each district. The data begins in 2011 and finish in 2021. Throughout this time, each district may have had 1 or more instances of renovation, happening in different years, eventually adding up. Some districts had 0 interventions.

Can I use a DID to study the effect of this treatment? The dependent variable is the housing costs in each district.

This is an example of my treatment data:

ID Number_of_interventions Year_of_each_intervantion 21 5 2015-16-19-20-21 11 3 2018-20-20-20 35 4 2018-18-19-20 59 4 2017-17-20-20 1 3 2018-19-20 22 3 2019-19-20 44 3 2015-15-19 57 3 2015-19-19 85 3 2015-16-18 49 2 2017-18 9 2 2018-19 14 2 2019-20 23 3 2016-18 31 2 2016-16 33 2 2017-20 42 2 2018-19 51 2 2019-19 56 2 2020-20 71 2 2016-17 76 2 2017-17 77 2 2014-17 2 1 2016 4 1 2018 5 1 2019 6 1 2013 73 0 74 0 75 0 78 0 80 0 81 0 83 0 84 0 86 0 87 0 88 0

[Daniël van Hemert]

Yes, in principle, if you reformat the data. However, your post alone does not answer whether a DID approach is most suitable.

See:

 
 ssc install did_multiplegt   help did_multiplegt

You should certainly test for pre-trends, it is likely that renovation decisions correlate with price trends.

[Raffaele Di Tria]

Dear Danile, thank you so much for your answer. I cannot view the code after "See:", perhaps it's my computer's problem?

[Daniël van Hemert]

Not sure why that is, but it states:

ssc install did_multiplegt
help did_multiplegt

You can also find more information on the related github: https://github.com/chaisemartinPacka...legt/tree/main

[Raffaele Di Tria]

Dear Daniel, thank you for this. I have been trying to use did_multiplegt, with the dyn mode, but the results have been weird. The graph output looks like the attached image. I think there may have been issues with the way I reformatted the data. I formatted them increasing the treatment variable by one when a new renovation was made in the district, and they look like this:

id_nil	year	treatment
1	2015	0
1	2016	0
1	2017	0
1	2018	1
1	2019	2
1	2020	3
1	2021	3
1	2022	3
1	2023	3

The command is as follows

 did_multiplegt (dyn) housingcost id year treatment,  controls(residents immigrants)

Do you have any idea why it is not working?

Thank you so much!