This paper is making empirical investigation of transmission of volatility shocks across the space and time, by using spatio-temporal form of GARCH (Generalized autoregressive conditional heteroscedastic) processes. Use of spatial-temporal weights allow both, past errors and past conditional variances in neighboring locations, to have differential impact on each current estimated variance at different transaction location. Hence, this spatio-temporal setting allow one to model both, the “shock” and “memory” of the conditional volatility. Estimated parameters using municipal housing transactions from Jonkoping, Sweden from 2015 through 2019 shows positive memory effect, along with positive shock and confirm the assumption of spatio-temporal persistence in the variance. Omitting the spatial GARCH element brings to the underestimation of the conditional variance. This finding favor the use of transactional spatial model vs aggregated model. Market for the houses has different spatio-temoral dynamics compared with the market for the apartments, showing stronger immediate spatial response to the shocks in the near neighborhood, but lower spatial memory (persistence) of the past close-by shocks.