two dummy variables interactions with specific condition

[Khati Zolfaghari]

Hello, have panel data of 171 countries from 1996-2019, and "https://www.statalist.org/forums/forum/general-stata-discussion/general/1675380-graph-for-evolution-variable-in-panel-form" is part of my data, and my baseline model is :

xtscc d0ltotalfertility l(0/2).pand_res_pt#(1-ae) l(0/2).pand_res_pt#non-ae l(1/1)d0ltotalfertility  i.year,fe

which

pand_res_pt

is dummy variables that get 1 in the time of recession and 0 otherwise, which is a shock in my model.

and "ltotalfertility" is the log of fertility rate

I wonder to see the analysis for advanced vs non-advanced with the interaction of the shock with a dummy – for total fertility rate, so first of all I generate the dummy variable "ae"=1 of an advanced economy and 0 otherwise.
and for the non-advanced also generate the non_adv=1 if non-advanced and 0 otherwise.

now Y=a+b1*x(1-adv) + b2*adv,

but in my regression when I add the interaction term, I did :

xtscc d0ltotalfertility l(0/2).pand_res_pt#(1-ae) l(0/2).pand_res_pt#ae l(1/1)d0ltotalfertility  i.year,fe

I received an error that

1 invalid name
r(198);

but when I estimate in the simple regression : xtscc d0ltotalfertility l(0/2).pand_res_pt l(0/2).pand_res_pt l(1/1)d0ltotalfertility i.year if ae==1 ,fe

I get the results:

 xtscc d0ltotalfertility l(0/2).pand_res_pt  l(1/1)d0ltotalfertility  i.year if ae==1,fe

Regression with Driscoll-Kraay standard errors   Number of obs     =       480
Method: Fixed-effects regression                 Number of groups  =        24
Group variable (i): ifscode                      F( 23,    19)     =      4.36
maximum lag: 2                                   Prob > F          =    0.0009
                                                 within R-squared  =    0.2858

-----------------------------------------------------------------------------------
                  |             Drisc/Kraay
d0ltotalfertility | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
------------------+----------------------------------------------------------------
      pand_res_pt |
              --. |  -.0118819   .0101403    -1.17   0.256    -.0331058    .0093421
              L1. |  -.0082734   .0045952    -1.80   0.088    -.0178913    .0013446
              L2. |  -.0246423   .0090083    -2.74   0.013     -.043497   -.0057877
                  |
d0ltotalfertility |
              L1. |   .0418261   .0946173     0.44   0.663    -.1562101    .2398624
                  |
             year |
            1996  |          0  (empty)
            1997  |          0  (omitted)
            1998  |   .0111177   .0003693    30.10   0.000     .0103447    .0118907
            1999  |   .0258679   .0010344    25.01   0.000     .0237028     .028033
            2000  |   .0422096   .0003334   126.62   0.000     .0415119    .0429074
            2001  |    .002944   .0019368     1.52   0.145    -.0011097    .0069977
            2002  |    .025964   .0017114    15.17   0.000     .0223821     .029546
            2003  |   .0301912   .0005511    54.78   0.000     .0290377    .0313447
            2004  |   .0344751    .000893    38.61   0.000     .0326061    .0363441
            2005  |   .0273988   .0011047    24.80   0.000     .0250867     .029711
            2006  |   .0459436   .0004741    96.91   0.000     .0449513    .0469359
            2007  |   .0372346    .002296    16.22   0.000     .0324291    .0420402
            2008  |   .0440433   .0015482    28.45   0.000      .040803    .0472836
            2009  |   .0220309   .0085197     2.59   0.018     .0041991    .0398628
            2010  |   .0364267   .0031583    11.53   0.000     .0298164    .0430371
            2011  |   .0287997    .006892     4.18   0.001     .0143746    .0432247
            2012  |   .0251388   .0016314    15.41   0.000     .0217242    .0285534
            2013  |  -.0046172   .0006605    -6.99   0.000    -.0059997   -.0032346
            2014  |   .0349674   .0031627    11.06   0.000     .0283479     .041587
            2015  |   .0220649    .000931    23.70   0.000     .0201162    .0240136
            2016  |   .0229065   .0003692    62.04   0.000     .0221337    .0236792
            2017  |          0  (omitted)
            2018  |          0  (omitted)
            2019  |          0  (omitted)
                  |
            _cons |   -.023716   .0001806  -131.34   0.000    -.0240939    -.023338
-----------------------------------------------------------------------------------

, I hope I could receive your advice and assistance so soon on how can I do the interaction in the model without generating manually before on the regression,


many thanks in advance for your valuable time and advice.

Best regards,

[Clyde Schechter]

Caveat: I am not familiar with the -xtscc- command, which is not part of official Stata*. But I don't think that matters. The problem here is a general one relating to factor-variable notation syntax. The things that appear in an interaction term in factor-variable notation must be variables, not expressions. (1-ae) is not a variable. That's the source of your problem.

In any case, your approach to this, even if it worked, is unnecessarily complicated. You have no need of several variables for your purpose. You just need to include the interaction with ae. So the code

xtscc d0ltotalfertility l(0/2).pand_res_pt#ae  l(1/1)d0ltotalfertility i.year,fe

is all you need (assuming it is otherwise correct, which, if it works like official Stata commands do, it should be).

*I am not claiming to be familiar with all official Stata commands.

[Khati Zolfaghari]

@Clyde Schechter thank you so much for your reply. I follow your advice, but I become a bit confused that in this case how can I interpret this because I could not use the (1-ae) as you mentioned to me it is not a variable. how can I see the difference between the advanced and non-advanced economies, and test them?

I am thinking about some test like:

y=a+b2*X*(1-D)+b3*X*D +c*D

b2=impct when the country is not advanced; D=0


b3=impact when the country is advanced; D=1

y=a+b+b1*D


bo+b1*D

bo= is impact when D=0 when country is not advanced=b2

bo+b1 when D=1 or advanced economies

b0=b2; bo+b1=b3

test bo+b1-bo=b1

if this kind of test is correct !!

. xtscc d0ltotalfertility l(0/2).pand_res_pt#ae  l(1/1)d0ltotalfertility i.year,fe

Regression with Driscoll-Kraay standard errors   Number of obs     =      3619
Method: Fixed-effects regression                 Number of groups  =       182
Group variable (i): ifscode                      F( 26,    19)     = 809493.49
maximum lag: 2                                   Prob > F          =    0.0000
                                                 within R-squared  =    0.1630

-----------------------------------------------------------------------------------
                  |             Drisc/Kraay
d0ltotalfertility | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
------------------+----------------------------------------------------------------
   pand_res_pt#ae |
             0 0  |          0  (empty)
             0 1  |          0  (omitted)
             1 0  |   .0026918   .0018903     1.42   0.171    -.0012646    .0066482
             1 1  |  -.0173334   .0021461    -8.08   0.000    -.0218253   -.0128416
                  |
 L.pand_res_pt#ae |
             0 0  |          0  (empty)
             0 1  |          0  (omitted)
             1 0  |  -.0081136   .0013466    -6.03   0.000    -.0109321   -.0052952
             1 1  |   -.002655   .0047199    -0.56   0.580    -.0125339     .007224
                  |
L2.pand_res_pt#ae |
             0 0  |          0  (empty)
             0 1  |          0  (omitted)
             1 0  |  -.0008307   .0018471    -0.45   0.658    -.0046967    .0030352
             1 1  |  -.0209225    .006022    -3.47   0.003    -.0335266   -.0083183
                  |
d0ltotalfertility |
              L1. |   .2662433   .0604578     4.40   0.000     .1397036     .392783
                  |
             year |
            1996  |          0  (empty)
            1997  |          0  (omitted)
            1998  |  -.0029064   .0008691    -3.34   0.003    -.0047256   -.0010873
            1999  |   .0019696   .0008357     2.36   0.029     .0002204    .0037188
            2000  |   .0052361    .000527     9.94   0.000     .0041331    .0063391
            2001  |  -.0061956   .0002488   -24.90   0.000    -.0067163   -.0056749
            2002  |   .0053032   .0008633     6.14   0.000     .0034964      .00711
            2003  |   .0065089   .0003361    19.37   0.000     .0058055    .0072124
            2004  |   .0093405   .0001266    73.76   0.000     .0090754    .0096056
            2005  |   .0059029   .0001135    51.99   0.000     .0056653    .0061406
            2006  |   .0123382   .0000582   211.94   0.000     .0122163      .01246
            2007  |   .0122363   .0003143    38.93   0.000     .0115784    .0128943
            2008  |    .013889   .0004023    34.52   0.000     .0130469     .014731
            2009  |   .0058794   .0007483     7.86   0.000     .0043131    .0074457
            2010  |   .0099586   .0004209    23.66   0.000     .0090776    .0108396
            2011  |   .0059901   .0010854     5.52   0.000     .0037184    .0082619
            2012  |   .0119376   .0002247    53.13   0.000     .0114674    .0124079
            2013  |    .000501   .0002738     1.83   0.083     -.000072    .0010741
            2014  |   .0128402   .0003622    35.45   0.000     .0120822    .0135982
            2015  |   .0058251   .0002527    23.06   0.000     .0052962    .0063539
            2016  |   .0071283   .0000588   121.20   0.000     .0070052    .0072514
            2017  |          0  (omitted)
            2018  |          0  (omitted)
            2019  |          0  (omitted)
                  |
            _cons |  -.0132229   .0005148   -25.69   0.000    -.0143003   -.0121454
-----------------------------------------------------------------------------------

Many thanks in advance for your valuable time and advice.

Best regards,

[Clyde Schechter]

margins pand_res_pt, dydx(ae)

will give you the contrasts between advanced and non-advanced economies for both shock and non-shock conditions. If you want to contrast the advanced and non-advanced economies without regard to (marginal distribution over) shock, then it's

margins, dydx(ae)

However, I also notice I made an important typo in the code I suggested in #2. It should be

xtscc d0ltotalfertility l(0/2).pand_res_pt##ae l(1/1)d0ltotalfertility i.year,fe

[Khati Zolfaghari]

@Clyde Schechter thanks for your reply. Sorry, if I ask because when I try to see the contrast between the advanced and non-advance I get an error. It could be becasue of the "xtscc" which I used in terms of the
"Driscoll-Kraay standard errors", but I am not sure?

. xtscc d0ltotalfertility l(0/2).pand_res_pt##ae l(1/1)d0ltotalfertility i.year,fe

Regression with Driscoll-Kraay standard errors   Number of obs     =      3619
Method: Fixed-effects regression                 Number of groups  =       182
Group variable (i): ifscode                      F( 26,    19)     = 809493.49
maximum lag: 2                                   Prob > F          =    0.0000
                                                 within R-squared  =    0.1630

-----------------------------------------------------------------------------------
                  |             Drisc/Kraay
d0ltotalfertility | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
------------------+----------------------------------------------------------------
      pand_res_pt |
               0  |          0  (empty)
               1  |   .0026918   .0018903     1.42   0.171    -.0012646    .0066482
                  |
    L.pand_res_pt |
               0  |          0  (empty)
               1  |  -.0081136   .0013466    -6.03   0.000    -.0109321   -.0052952
                  |
   L2.pand_res_pt |
               0  |          0  (empty)
               1  |  -.0008307   .0018471    -0.45   0.658    -.0046967    .0030352
                  |
               ae |
               0  |          0  (empty)
               1  |          0  (omitted)
                  |
   pand_res_pt#ae |
             0 0  |          0  (empty)
             0 1  |          0  (empty)
             1 0  |          0  (empty)
             1 1  |  -.0200252   .0030858    -6.49   0.000     -.026484   -.0135665
                  |
 L.pand_res_pt#ae |
             0 0  |          0  (empty)
             0 1  |          0  (empty)
             1 0  |          0  (empty)
             1 1  |   .0054587   .0056378     0.97   0.345    -.0063413    .0172586
                  |
L2.pand_res_pt#ae |
             0 0  |          0  (empty)
             0 1  |          0  (empty)
             1 0  |          0  (empty)
             1 1  |  -.0200917   .0052764    -3.81   0.001    -.0311353   -.0090481
                  |
d0ltotalfertility |
              L1. |   .2662433   .0604578     4.40   0.000     .1397036     .392783
                  |
             year |
            1996  |          0  (empty)
            1997  |          0  (omitted)
            1998  |  -.0029064   .0008691    -3.34   0.003    -.0047256   -.0010873
            1999  |   .0019696   .0008357     2.36   0.029     .0002204    .0037188
            2000  |   .0052361    .000527     9.94   0.000     .0041331    .0063391
            2001  |  -.0061956   .0002488   -24.90   0.000    -.0067163   -.0056749
            2002  |   .0053032   .0008633     6.14   0.000     .0034964      .00711
            2003  |   .0065089   .0003361    19.37   0.000     .0058055    .0072124
            2004  |   .0093405   .0001266    73.76   0.000     .0090754    .0096056
            2005  |   .0059029   .0001135    51.99   0.000     .0056653    .0061406
            2006  |   .0123382   .0000582   211.94   0.000     .0122163      .01246
            2007  |   .0122363   .0003143    38.93   0.000     .0115784    .0128943
            2008  |    .013889   .0004023    34.52   0.000     .0130469     .014731
            2009  |   .0058794   .0007483     7.86   0.000     .0043131    .0074457
            2010  |   .0099586   .0004209    23.66   0.000     .0090776    .0108396
            2011  |   .0059901   .0010854     5.52   0.000     .0037184    .0082619
            2012  |   .0119376   .0002247    53.13   0.000     .0114674    .0124079
            2013  |    .000501   .0002738     1.83   0.083     -.000072    .0010741
            2014  |   .0128402   .0003622    35.45   0.000     .0120822    .0135982
            2015  |   .0058251   .0002527    23.06   0.000     .0052962    .0063539
            2016  |   .0071283   .0000588   121.20   0.000     .0070052    .0072514
            2017  |          0  (omitted)
            2018  |          0  (omitted)
            2019  |          0  (omitted)
                  |
            _cons |  -.0132229   .0005148   -25.69   0.000    -.0143003   -.0121454
-----------------------------------------------------------------------------------

and my error:

. margins pand_res_pt, dydx(ae) warning: cannot perform check for estimable functions. default prediction is a function of possibly stochastic quantities other than e(b) r(498);

or even without cosidering the shock as you recomended me:

. margins, dydx(ae) warning: cannot perform check for estimable functions. default prediction is a function of possibly stochastic quantities other than e(b) r(498);

Many thanks in advance for your valuable time and advice. Best regards,

[Clyde Schechter]

Not being familiar with -xtscc-, I don't think I can advise you further here. Perhaps others who have been following the thread can help and will join in.