Spatial Econometrics

Spatial data analysis is a multidisciplinary analytical method that considers not only the content-based characteristics of observations but also their relationships with each other based on geographic location. This method, widely used across fields ranging from economics to geography and from agriculture to urban planning, distinguishes itself from traditional analyses by directly incorporating space into the analytical process. The primary goal in spatial data analysis is to achieve more consistent and realistic results by taking into account the spatial proximity relationships among units.

Spatial data include not only the characteristics of observed units but also their positions in space. Therefore, spatial data can be defined as a sequence of random variables ordered by location information. Unlike time series, spatial dependence is not limited to past observations; a given observation unit is influenced by other units that are spatially close. This phenomenon is known as spatial dependence.

Spatial data analysis has two fundamental challenges: spatial dependence and spatial heterogeneity.

• Spatial dependence is based on Tobler’s First Law of Geography: “Everything is related to everything else, but near things are more related than distant things.” This law implies that neighboring regions are more likely to exhibit similar characteristics (for example, high unemployment rates may also be observed in adjacent regions).
• Spatial heterogeneity arises when the parameters of a model are not constant across space. In other words, the same model may have different effects in different regions.

One of the fundamental building blocks of spatial data analysis is the spatial weight matrix. This matrix represents the neighborhood relationships among observations. There are two main types:

• Contiguity-based: Shared boundaries are considered (for example, if two provinces share a border, the corresponding matrix cell is assigned a value of 1).
• Distance-based: Weights are assigned according to physical distance.

Through this matrix, a change in a specific observation becomes part of a structure that affects not only that observation but also its neighbors.

Spatial regression models are statistical tools that incorporate spatial dependence into the modeling process. The most common models are:

• SAR (Spatial Autoregressive Model): Assumes that the dependent variable is influenced by values in neighboring regions.
• SEM (Spatial Error Model): Captures situations where error terms in the model are correlated across neighboring units.
• SDM (Spatial Durbin Model): Simultaneously accounts for spatial effects of both dependent and independent variables.

These models are selected and evaluated using diagnostic tests such as Moran’s I and LM tests.

Spatial data analysis is applied across numerous fields, from social sciences to engineering:

• Agriculture: Analyses can be conducted on the effects of vegetable production areas on neighboring regions (for example, if production increases in adjacent provinces, a similar upward trend may be observed in those provinces).
• Regional Development: Spillover effects can be identified in analyses of economic growth between provinces.
• Social Sciences: The influence of social norms, neighborhood effects, and reference groups on individuals can be analyzed using spatial modeling.
• Integration with Data Envelopment Analysis and Principal Component Analysis: As in the study by Ceren Yaman Yılmaz, results obtained from DEA and PCA can be integrated into spatial regression models to enable more in-depth analyses.