Generating a dummy variable if a variable increases over time

[Timo de Haan]

Dear all,

I am trying to create a dummy variable that indicates whether a variable increases over time or not. I have a variable 'perrank' that indicates the place of an individual in a wealth distribution. For example, ID=101 in year = 2015 with perrank = 86, indicates that an individual in 2015 belonged to the top 14 percent. Now I want to calculate whether that person's place in the wealth distriution increases over time. I want to create a dummy that says 1 if increase over time and 0 if not. So if ID =101 in year = 2016 with perrank = 87, the dummy should have a 1, because the perrank variable increased. Preferably, I would be able to change the degree of change. For example, whether the increase is one percentile or two percentiles.

I have seen this forum (https://www.statalist.org/forums/for...r-time-periods) but I could not write the code to solve my problem. It is probably something like ' bysort id (year), sort: gen changed = .............'. I have panel data. The ultimate goal is to calculate the probability of rising in the wealth distribution after receiving inheritance.

Thank you in advance.

Best,

Timo

[Nick Cox]

bysort id (year) : gen decrease = sum(cond(_n == 1, 0, perrank < perrank[_n-1]))
by id : replace decrease = decrease[_N]

will count decreases. If staying the same is what you want to record too change < to <=

Then decrease equal to 0 means always increasing or stable.

Warning: not tested (you didn't give example data). Missings may need more care than this.

[Timo de Haan]

Thank you for the response Nick. The code is, however, not giving the results I would like. Using 'tab' it shows that it adds the values:

. tab decrease

decrease | Freq. Percent Cum.
------------+-----------------------------------
0 | 10,694 15.54 15.54
1 | 12,905 18.76 34.30
2 | 10,655 15.49 49.78
3 | 6,639 9.65 59.43
4 | 4,279 6.22 65.65
5 | 3,304 4.80 70.45
6 | 2,896 4.21 74.66
7 | 2,787 4.05 78.71
8 | 2,640 3.84 82.55
9 | 2,856 4.15 86.70
10 | 2,367 3.44 90.14
11 | 2,070 3.01 93.15
12 | 1,937 2.82 95.96
13 | 1,041 1.51 97.48
14 | 847 1.23 98.71
15 | 537 0.78 99.49
16 | 215 0.31 99.80
17 | 82 0.12 99.92
18 | 56 0.08 100.00
------------+-----------------------------------
Total | 68,807 100.00

Data browsing shows that the 'dummy' is already giving numbers as 2 or more in the first year. Could you take a second look at the code? I want to create a dummy that says 1 if percentile ranks over time increases with 1% for an individual. Thus, if perrank increases from 87 to 89 from the year 2010 to 2011, the dummy should say 1. If from 2011 to 2012 it increases again from 89 to 91 it should say another time 1.

Here is example data. It is a unbalanced panel. I have added a column 'increase' to show what I mean. There is no missingness in the data. One problem I now recognize is that I am looking at changes over time. This has problems for the first year of each individual: should it be a zero or a '.'? Any advice?

id	year	twealth_corrected	perrank	increase
101	1993	0	0.230083	./0
101	1994	0	0.272502	1
101	1995	0	0.213183	0
102	1993	68849.09	0.76121	./0
102	1994	-15209	0.01407	0
102	1995	52459.7	0.691973	1
201	1993	117970.1	0.861559	./0
201	1994	58939.64	0.865301	1
202	1993	4718.422	0.447323	./0
202	1994	4698.9	0.512932	1
301	1993	-7513.7	0.019373	./0
301	1994	-3197.98	0.038899	1
401	1993	106315.5	0.84458	./0
401	1994	23417.32	0.742189	0
401	1995	165937	0.912987	1
401	1996	95108	0.772455	0
402	1993	0	0.230083	./0
402	1994	908	0.370991	0
402	1995	908	0.30955	0
402	1996	1035.12	0.291667	0
501	1993	0	0.230083	./0
501	1994	0	0.272502	1
501	1995	-22273.2	0.007179	0

Thank you in advance.

[Timo de Haan]

I did not receive a response yet, while the response is often fast. Is there anything unclear? I would love to clear things up