Before describing my problem in detail, here is my specific question: Can I use spregress without introducing fixed effects/ random effects at the id level when my goal is to simply use spregress's ability to address the issue of spatial correlation across observations?

Now for some details. First, the research question I am interested in: How does municipal-level political competition affect the funding level of a pension plan run by that municipality?

To simplify the structure of my data is as follows: I have data on a municipal-level measure of political competition and funding level of various pension plans run by these municipalities over time. Each municipality belongs to a county and for a number of reasons, my baseline specification includes county fixed effects.

Thus, this is my (simplified) baseline regression:
Code:
xi: areg meanactfundratio lmeanactfundratio political_competitiveness7 i.period, absorb(county)
where meanactfundratio is the funded ratio; lmeanactfundratio is the funded ratio lagged one period, and political_competitiveness7 is the measure of political competition. I also introduce period dummies.

Now I want to address the spatial correlation between pension plans. I believe one somewhat unsatisfactory approach of doing that is to cluster standard errors at the county level, which I have done. Thus I have run:
Code:
xi: areg meanactfundratio lmeanactfundratio political_competitiveness7 i.period, absorb(county) cluster(county)
Additionally I would like to explore the use of the sp-suite of commands because those are specifically designed to address spatial correlation. My challenge here however is that while the data are a panel of multiple observations per municipality, I am not interested in introducing municipality fixed/ random effects. My specific question is: Can I use spregress to address spatial correlation without introducing either municipality fixed effects/ random effects (but including county fixed effects) when the underlying data are at the municipality-period level?

While I am able to run:
Code:
xi: spxtregress meanactfundratio lmeanactfundratio political_competitiveness7 i.period, re errorlag(W)
this regression uses random effects at the municipality level -- something that wasn't a part of the baseline specification presented earlier.

I am however unable to run:
Code:
 xi: spregress meanactfundratio lmeanactfundratio political_competitiveness7 i.period i.county, ml errorlag(W)
This is the error message I get:
(2074 observations)
(2074 observations (places) used)
(weighting matrix defines 1037 places)
estimation sample defines places not in weighting matrix
You must specify if or in to restrict the estimation sample to the places in the weighting matrix or use a different weighting matrix.

Code:
* Example generated by -dataex-. To install: ssc install dataex
clear
input float(meanactfundratio lmeanactfundratio political_competitiveness7 period) str6 county float municipality_id_num str8 municipality_id
 90.79351 38.034103   -.05151126 1996 "01"  1 "010193"
103.80653  90.79351    -.1492273 2006 "01"  1 "010193"
 43.48037  21.57011    -.0984934 1996 "01"  2 "010215"
 48.72267  43.48037   -.22404087 2006 "01"  2 "010215"
126.14955 346.06445   -.05798948 1996 "01"  3 "010245"
102.04782 126.14955   -.12199911 2006 "01"  3 "010245"
155.18576  358.4015   -.05018273 1996 "01"  4 "010275"
 126.8356 155.18576   -.13270164 2006 "01"  4 "010275"
 88.97825  83.60065   -.06226122 1996 "01"  5 "010303"
 84.76244  88.97825    -.1883972 2006 "01"  5 "010303"
113.68117 100.31748 -.0006153584 1996 "01"  6 "010453"
100.70793 113.68117  -.036280274 2006 "01"  6 "010453"
 369.6445 216.80074   -.07789397 1996 "01"  7 "010515"
537.20825  369.6445    -.1784345 2006 "01"  7 "010515"
128.88698 16.939756   -.04999974 1996 "01"  8 "010635"
 97.73939 128.88698   -.16537768 2006 "01"  8 "010635"
 85.15027 121.13144  -.026643127 1996 "01"  9 "010663"
104.20557  85.15027     -.180543 2006 "01"  9 "010663"
 102.5975 146.34023   -.14475524 1996 "01" 10 "010693"
104.22276  102.5975   -.04251865 2006 "01" 10 "010693"
end
If instead of using Stata's sp suite of commands, I could use another command to address this spatial correlation, I would be keen to hear of such an approach. Lastly, I am using Stata version 16.

Thank you for any and all suggestions.
Sutirtha Bagchi