Problem on "statsby" function

[Cedric Hung]

Dear all,

I am doing a banking competition analysis of EU28 countries with the data for the year 2011-2016. The dataset is unbalanced. I have to calculate a few competition measures by running a Stochastic Frontier Analysis and few fixed effect regression for each country and year.

Then I would like to save the results of coefficients for each country and year. I have tried the following command but what I get is just a dataset with all missing data. Can anyone suggest how can I do it properly? I have run the following command for two of the measures:


1. Boone indicator

statsby bi=_b[lac], by(year country) clear nodots: xtreg roa lac leqast lta lloanast ldeptf, fe

2. H-statistic

statsby _b[lp1] _b[lp2] _b[lp3], by(year country) clear nodots : xtreg ltrta lp1 lp2 lp3 leqast lta lloanast ldeptf, fe

Also, how can I save down v_it of the error term from Stochastic Frontiere Analysis?

Many thanks,
Cedric

[Nick Cox]

You're asking for a separate xtreg for each country-year combination. Each such combination defines a single observation in the data if it's been xtset. You can't fit a model with several predictors to a single observation. So the syntax is legal, but the request makes no statistical or economic sense and the result is inevitable.

 xtreg roa lac leqast lta lloanast ldeptf, fe

may make sense by itself (it seems that you have at most 28 x 6 observations to play with). A regression for each country is likely to be out of the question; you don't have enough data.

(NB: statsby is a command, not a function.)

Sorry, but I don't know anything much about stochastic frontier analysis.

[Cedric Hung]

Dear Nick,

Thanks for the response. Here is the table showing the numbers of observations I have for each country:

Country	Freq.	Percent	Cum.
Austria	2,367	35	35
Belgium	115	1.7	36.71
Bulgaria	89	1.32	38.02
Croatia	141	2.09	40.11
Cyprus	103	1.52	41.63
Czech	84	1.24	42.87
Denmark	283	4.19	47.06
Estonia	30	0.44	47.5
Finland	102	1.51	49.01
France	502	7.42	56.43
Greece	37	0.55	56.98
Hungary	84	1.24	58.22
Ireland	29	0.43	58.65
Italy	514	7.6	66.25
Latvia	61	0.9	67.15
Lithuania	38	0.56	67.72
Luxembourg	267	3.95	71.67
Malta	42	0.62	72.29
Netherlands	122	1.8	74.09
Poland	273	4.04	78.13
Portugal	326	4.82	82.95
Romania	99	1.46	84.41
Slovakia	54	0.8	85.21
Slovenia	44	0.65	85.86
Spain	222	3.28	89.15
Sweden	318	4.7	93.85
United Kingdom	416	6.15	100
Total	6,762	100

			year
Country	2011	2012	2013	2014	2015	2016	Total
Austria	368	368	434	368	420	409	2,367
Belgium	13	15	20	20	23	24	115
Bulgaria	9	9	18	18	18	17	89
Croatia	19	19	26	26	26	25	141
Cyprus	4	4	13	26	27	29	103
Czech	6	6	18	18	18	18	84
Denmark	32	31	53	53	57	57	283
Estonia	2	2	6	6	7	7	30
Finland	3	3	15	24	32	25	102
France	27	29	104	114	115	113	502
Greece	6	6	7	6	6	6	37
Hungary	6	6	18	18	18	18	84
Ireland	3	3	6	6	5	6	29
Italy	74	76	87	92	91	94	514
Latvia	1	1	15	15	14	15	61
Lithuania	5	5	8	8	6	6	38
Luxembourg	25	27	54	57	54	50	267
Malta	4	4	8	9	9	8	42
Netherlands	13	14	23	24	25	23	122
Poland	16	18	50	55	61	73	273
Portugal	4	4	20	95	99	104	326
Romania	8	8	19	21	21	22	99
Slovakia	6	6	11	11	11	9	54
Slovenia	7	7	7	8	7	8	44
Spain	12	13	48	50	50	49	222
Sweden	7	7	71	75	78	80	318
United Kingdom	30	30	83	89	93	91	416
Total	710	721	1,242	1,312	1,391	1,386	6,762

.

cannot be run which I only get a dataset of all missing data when I use the "statsby" to save the coefficients?


Many thanks,
Cedric

[Nick Cox]

So, you have multiple observations for each year for many countries, but by no means all. What then was your xtset command?