I would like to run a regression discontinuity in Time in R. First, I would like to visualize the regression using the rdplot function. The code I have used you can find below. With c I should specify the cutoff point based on the running variable. However, the rd package is designed to use a normal continuous variable and not a date as the running variable. So, I get the error message Error in `-.Date`(x_max, c) : can only subtract from "Date" objects. I have the same problem when I use the rdrobust command where I also have to specify the cutoff.
The code for the rdplot I used is:
rdplot(
df$Date, df$dependentvariable,
c = as.Date(as.character("20161231"), format = "%Y%m%d"),
nbins = 4
)
The code for the rdrobust I used is:
rdplot(
df$Date, df$dependentvariable,
c = as.Date(as.character("20161231"), format = "%Y%m%d")
)
Does anyone have any idea whether I can use the rdrobust and rdplot command for regression discontinuity in time? If yes, how do I have to adjust the code? Are there any other packages/commands that I can use to run a regression discontinuity in time or are there any workarounds?
Edit due to first answer asking for more information:
Thank you for the answer. The data set is confidential and very large, so I can not share it here. However, I created an exemplary, shorter data set with the same structure as mine. It has the following structure:
No, Company, IPO Date, IPO mechanism, IR (Underpricing), firm age (Age) at listing
1,Company 1, 2016-11-01, 0, 10.3%, 5.9
2,Company 2, 2016-11-10, 0, 5.2%, 10.3
3,Company 3, 2016-11-12, 0, -1.1%, 3.4
4,Company 4, 2016-12-01, 0, 6.5%, 4.3
5,Company 5, 2016-12-09, 0, 8.3%, 6.2
6,Company 6, 2016-12-16, 0, 7.2%, 10.5
7,Company 7, 2016-12-23, 0, 4.1%, 7.4
8,Company 8, 2016-12-28, 0, -6.5%, 7.7
9,Company 9, 2017-01-01, 1, 1.3%, 7.0
10,Company 10, 2017-01-10, 1, 2.2%, 3.4
11,Company 11, 2017-01-12, 1, -4.1%, 5.5
12,Company 12, 2017-01-01, 1, -10.5%, 9.2
13,Company 13, 2017-01-09, 1, 6.2%, 7.6
14,Company 14, 2017-01-16, 1, 2.5%, 11.5
15,Company 15, 2017-02-23, 1, 4.1%, 7.4
16,Company 16, 2017-02-28, 1, -6.5%, 7.7
Note: Various other covariates than age are included, but not shown in the exemplary data set.
A change in the regulatory framework at the 2016-12-31 changed the selling mechanism of the IPO. I want to measure the impact of the regulatory change on underpricing. I know this is not a typical RD design but it should nevertheless be applicable. I am also aware that I have to be very carefully with the definition of the bandwidth.
So what I want to do: I would like to run a discontinuity regression in time where the cutoff is the date when the new regulation was implemented. The running variable is therefore time and the independent variable underpricing. In addition, I would like to include various company and IPO specific covariates like firm age or expected issue size which have an impact on underpricing according to prior research.
Further, I would like to plot the relationship between time and underpricing by grouping the data into points. Similar to what the rdplot function is doing.
So what I did you can see in the original post above
I think that there are mainly two problems with the rdplot and rdrobust function from the rdrobust package: 1. It does not recognize the cutoff as a date (the error message .... appears) 2. I cannot include covariates in the equation
Maybe there is also a workaround by splitting the data set at the cut off and afterwards run two separated regressions. However, for the plot I don't know how I would have to do it.
As I am not an R expert, I am totally lost on how to run this regression. Therefore, thank you very much already beforehand!
