plm/splm in R - fixed effects panel models with time invariant variables

a bit new to working with panel data models. i’m estimating the following model for crime rate:

tot_crime <- log_rt_crime ~ ndvi + log_disad_inde + log_density

where ndvi is a time-variant continuous variable, and log_disad_inde and log_density are time-invariant continuous variables. in brief, here is the data structure:

geoid           log_density     log_disad_inde  time    ndvi    log_rt_crime
170310101001    0.6633400        0.9688232      1   0.21318862   2.213387
170310101001    0.6633400        0.9688232      2   0.46648943   1.961499
170310101001    0.6633400        0.9688232      3   0.39554212   2.725013
170310101001    0.6633400        0.9688232      4   0.14324121   2.414422
170310101002    0.8935242        1.1857459      1   0.13986452   3.758296
170310101002    0.8935242        1.1857459      2   0.26278001   3.559268
170310101002    0.8935242        1.1857459      3   0.20211675   3.646066
170310101002    0.8935242        1.1857459      4   0.09324110   3.390321

when i estimate the individual fixed effects within estimator without spatial effects:

summary(FEwithin_nosp_crime <- plm(tot_crime, data = final_long_panel, model = "within", effect = "individual")),

R understandably only estimates my time-variant variable, ndvi, and drops log_density and log_disad_inde. however, when i account for spatial dependency,

summary(FEwithin_SAR_crime <- spml(tot_crime, data = final_long_panel, listw = queen, lag = TRUE, model = "within", effect = "individual", spatial.error="none")),

i do get estimates for my time-invariant variables. why is this the case? cheers.

Note: above model is the spatial lag model; i also am getting estimates for time-invariant variables when i estimate the spatial error model and other spatial panel specifications.