Problem with generating a variable from the temporal difference of another variable

[Tilman Schaefer]

Dear Community,

I am looking for a solution to my problem; I am using Stata 17.0 and am still very inexperienced in using the program. I am investigating certain factors influencing foreign aid's effectiveness for democratisation. To measure the level of democratization, I am using the PolityV Democracy Index, which shows a 21-point scale from autocratic (-10 to -6), anocratic (-5 to 5) and democratic (6-10) for countries with more than 500k citizens. My Dataset includes 3520 observations and, for now, gathers the following variables:

year - For the respective year, foreign aid flows were transferred to the recipient country. It is important to note that this did not happen yearly or that no data exists, as seen in the data sample.

recipient - The recipient country which received the monetary aid

aid_disbursement - The sum of all foreign aid flows a country received in a given year

democratization - The PolityV value attributed to the recipient country in the respective year

* Example generated by -dataex-. For more info, type help dataex
clear
input int year str56 recipient double aid_disbursement byte democratization
1990 "Afghanistan"   11712867 -8
1994 "Afghanistan"    1982707  0
1995 "Afghanistan"     139271  0
1996 "Afghanistan"      47409 -7
1998 "Afghanistan"     606751 -7
1999 "Afghanistan"    8447426 -7
2000 "Afghanistan"    7188738 -7
1992 "Albania"         586798  5
1993 "Albania"        6648893  5
1994 "Albania"        6268677  5
1995 "Albania"       23700096  5
1996 "Albania"        3200390  0
1997 "Albania"       23551309  5
1998 "Albania"       46664735  5
1999 "Albania"      238256793  5
2000 "Albania"       75143493  5
2001 "Albania"       47809639  5
2002 "Albania"       39137869  7
2003 "Albania"       79564698  7
2004 "Albania"       91081393  7
2005 "Albania"       82393557  9
2006 "Albania"       71182649  9
2007 "Albania"       83653257  9
2008 "Albania"       53898372  9
2009 "Albania"       59314383  9
2010 "Albania"       60401083  9
2011 "Albania"       39079953  9
2012 "Albania"       35262134  9
2013 "Albania"       75343224  9
1988 "Algeria"        6600713 -9
1989 "Algeria"      562354142 -2
1990 "Algeria"      391043394 -2
1992 "Algeria"       49387504 -7
1996 "Algeria"        5910180 -3
1997 "Algeria"         231739 -3
1998 "Algeria"        1258876 -3
1999 "Algeria"       13286871 -3
2000 "Algeria"       26016808 -3
2001 "Algeria"       47089893 -3
2002 "Algeria"        5144818 -3
2003 "Algeria"       45872711 -3
2004 "Algeria"       28078596  2
2005 "Algeria"       36738862  2
2006 "Algeria"       35723382  2
2007 "Algeria"       36044550  2
2008 "Algeria"        5728497  2
2009 "Algeria"        5926904  2
2010 "Algeria"       47974327  2
2011 "Algeria"        7062412  2
2012 "Algeria"       46949981  2
2013 "Algeria"       33267065  2
1984 "Angola"        11679797 -7
1986 "Angola"          157608 -7
1987 "Angola"           35793 -7
1988 "Angola"         1146468 -7
1989 "Angola"         5479661 -7
1990 "Angola"         3741361 -7
1991 "Angola"        36748934 -4
1992 "Angola"        26491601  0
1993 "Angola"         3714677 -1
1994 "Angola"          335855 -2
1995 "Angola"         2542596 -3
1996 "Angola"         2266317 -4
1997 "Angola"         4024587 -6
1998 "Angola"         2809307 -6
1999 "Angola"        23829711 -6
2000 "Angola"        19344245 -6
2001 "Angola"        19630813 -6
2002 "Angola"        24266683 -2
2003 "Angola"        41423017 -2
2004 "Angola"        27950980 -2
2005 "Angola"        21651019 -2
2006 "Angola"        87116433 -2
2007 "Angola"        67011834 -2
2008 "Angola"        25385743 -2
2009 "Angola"        49822128 -2
2010 "Angola"        39433785 -2
2011 "Angola"        14169800 -2
2012 "Angola"        20557775 -2
2013 "Angola"        30708579 -2
1971 "Argentina"       288538 -9
1984 "Argentina"       351913  8
1985 "Argentina"       168412  8
1986 "Argentina"     41148359  8
1988 "Argentina"       339281  8
1989 "Argentina"     11864703  7
1991 "Argentina"   1378206951  7
1992 "Argentina"     36502594  7
1993 "Argentina"       271856  7
1994 "Argentina"     69927153  7
1995 "Argentina"   1004889492  7
1996 "Argentina"    528379428  7
1997 "Argentina"    485531891  7
1998 "Argentina"    193236104  7
1999 "Argentina"      2555859  8
2000 "Argentina"    966701678  8
2001 "Argentina"    832062478  8
2002 "Argentina"      1694748  8
2003 "Argentina"      2237610  8
2004 "Argentina"     42188708  8
end

To determine the degree of democratization before and after receiving foreign aid flows, a new variable, regime change, is to be generated. Since regime changes happen very slowly, a period of five years should be considered.

This new variable should be measured as a change in the country's PolityV score five years after receiving foreign aid. To determine the extent of regime change, I planned to subtract the country's PolityV score at the time of aid allocation from its PolityV score five years later.

My problem is that I do not know how to proceed here to generate the required variable. Likewise, for many observations, data do not exist for every year, which is why the time span of five years cannot always be considered. I am happy for any advice on commands or literature I can have a look at.

Many thanks for considering my request
Tilman

[Clyde Schechter]

Perhaps I am misunderstanding what you want, but it seems that what you ask cannot be done with this data. The problem is that there are years where a country receives aid, but there is no information in the data about that country 5 years later. Examples are Afghanistan in 1996, 1998, 1999, and 2000, or Angola 2009, 2010, 2011, 2012, and 2013, to mention just a few of the 29 instances in the example data. So there is no way to identify what the democratization index 5 years after aid would be.

I think you need to get yearly data about the democratization index and add that to your data set, probably as more observations containing the PolityV score for those years, and with the aid_disbursement variable set to missing or zero. If that is not a possibility, you need to decide on a rule for how to handle the situation where no PolityV score is available 5 years later, a rule that does minimal damage to achieving your research goals.

[Tilman Schaefer]

Thank you very much for your prompt reply, Mr Schechter! You have understood my request correctly; unfortunately, as far as I know, there are no data available for the missing years for the PolityV index that also have foreign aid flows available in the same year. I can't think of a rule on how to work around this problem in the short term, but as soon as I know such a solution, I will immediately update this thread.

However, given the relatively high number of observations, I would be satisfied to include only the values for which data exist five years later in my analysis and omit those for which this is not possible and not consider them further. I think this leaves a decent number of observations to apply statistical models.

How would this be done with Stata? Omit all values that do not fulfil the requirement of the 5-year span and create the variable regime-change only for those values for which data also exist five years after aid disbursement?

Thanks
Tilman

[Clyde Schechter]

encode recipient, gen(country)
xtset country year
gen regime_change = F5.democratization - democratization

there are no data available for the missing years for the PolityV index that also have foreign aid flows available in the same year

That may be, but if there are data for the missing years for just PolityV, even without foreign aid data, adding that to the data set would enable you to calculate regime_change for many of the values where it is currently not knowable.