Criminological research is moving towards the study of small geographic areas. Crime and crime perceptions are influenced by environmental features and contextual conditions that are more common in some places than others, and therefore these are unequally distributed in space. By visualising criminological phenomena with maps at small area level, researchers are able to examine their immediate environmental explanations, and police forces can design targeted strategies to reduce crime and increase public safety. The two main sources of data for mapping crime are police records and surveys, and crime perceptions are mainly recorded by surveys. Although police-recorded crimes can be used for crime mapping, these suffer from a high risk of bias arising from victims underreporting. Victimisation surveys record information about unreported crimes, fear of crime and attitudes towards policing. However, surveys tend to be designed to record representative samples for large geographies, and small areas suffer from small sample sizes. Small samples do not allow for direct estimates of adequate precision. In order to produce reliable small area estimates of survey-recorded crime and perceptions about crime, small area estimation techniques introduce models to borrow strength across related areas. Small area estimators can incorporate spatially and temporally correlated random effects to increase the estimates reliability. The primary goal of this thesis is to bridge the gap between criminology and small area estimation, by providing a framework of theory, simulation experiments and applications for the use of small area estimation in criminological research. This is an alternative format thesis (by publications) including four papers framed between an introduction, literature review and conclusions. The first chapters present a discussion about the move in criminology towards the study of micro places, as well as an introduction to the small area estimation methods used in this dissertation (i.e. Empirical Best Linear Unbiased Predictor (EBLUP) based on Fay-Herriot model, Spatial EBLUP (SEBLUP), Rao-Yu model and Spatial-temporal EBLUP). The first paper provides a simulational assessment of the SEBLUP under different scenarios of number of areas and spatial autocorrelation, and produces estimates of confidence in policing at a ward level in London. The second paper produces estimates of worry about crime (burglary and violence) at a regional level in Europe and examines its predictors. The third paper produces estimates of perceived neighbourhood disorder in Manchester. The fourth paper presents estimates of crimes unknown to police (a measure of dark figure of crime) at neighbourhood and local level in England and Wales. Substantive and methodological theory and exemplar studies are integrated to show the utility of applying small area estimation to analyse some topics of interest in criminology. By expanding the body of research using small area estimation in criminological research, these methods may become a core tool for crime analysts and geographic criminologists.