The interview was conducted in Paris by Leonidas Doukissas during the 20th World Conference of the Spatial Econometrics Association.
Giuseppe Arbia is an Italian statistician internationally recognised as one of the leading figures in spatial statistics and spatial econometrics. He is Professor of Economic Statistics at Università Cattolica del Sacro Cuore and holds a PhD from the University of Cambridge. His work has made a major contribution to bridging statistical theory, econometric methodology, and the analysis of spatial data.
His extensive research record includes eight books and more than 100 scientific articles addressing spatial dependence and heterogeneity, regional convergence, the spatial concentration of economic activity, and spatial microeconometrics. Among his best-known works is A Primer for Spatial Econometrics, a widely used introduction to the field whose second edition includes applications in R, Stata, and Python.
Arbia has also played a central institutional role in the development of the discipline. He has served as Chairman of the Spatial Econometrics Association since 2006 and has been associated with major initiatives including the Spatial Econometrics Advanced Institute and the Journal of Spatial Econometrics.
In the conversation that follows, he reflects on his personal route into spatial analysis and explains why data cannot be considered independently of the places to which they refer. The discussion addresses the foundations of the field, the contribution of the Spatial Econometrics Association, the importance of R and Python, and the major future challenge of defining spatial proximity more objectively.
Giuseppe Arbia’s academic profile
From maps to spatial analysis
Leonidas Doukissas: Professor Arbia, your work has been central to the development of spatial statistics and spatial econometrics. How did you first become interested in the role of space in statistical and econometric analysis?
Giuseppe Arbia: To tell you the truth, my interest began because I had been fascinated by maps since childhood. I was always looking at them. I enjoy walking in the mountains and consulting maps, as well as looking at them while travelling by car. I therefore became familiar with reading maps from an early age.
When I graduated in 1981, I was looking for a new subject to which I could devote my research. I felt that all the topics I had been taught had reached a stage where very little new could be said.
At that time, time-series analysis was becoming very popular. It may be difficult to believe today, but in the 1980s most economics books contained only a few references to time series, let alone to space. I therefore began studying time-series analysis and autoregressive models. I then came across Space-Time Series Analysis, a book by Bob Bennett, a geographer at Cambridge.
I wrote him a handwritten letter—there was no internet at the time—and he replied and invited me to visit him in Cambridge. I stayed there for one year, after which he suggested that I remain for another two years to complete a PhD. That is where everything began.
Why does space matter?
Leonidas Doukissas: Why does space matter in statistical and econometric analysis?
Giuseppe Arbia: Everything happens in space and time. All the economic decisions made by economic agents are influenced by the environment in which people live. When you decide to go to a restaurant or to buy a house, every economic decision must take space into account because it involves moving from one place to another. The dynamics of geography are therefore of paramount importance in economics.
Spatial statistics and spatial econometrics
Leonidas Doukissas: How would you explain the relationship between spatial statistics and spatial econometrics to young researchers?
Giuseppe Arbia: That is a good question because the boundary between the two disciplines is not entirely clear. Generally speaking, spatial statistics is broader. It includes all the methods used to analyse data distributed in space.
Spatial econometrics, as an application of spatial statistics to economics, usually assumes that there is a theoretical model that we want to test. In spatial statistics, by contrast, a theoretical model is not always necessary.
There is a narrow definition of spatial econometrics that essentially refers to linear regression models applied to spatial data. However, when we created the Spatial Econometrics Association in 2006, we adopted a broader definition in its bylaws: all methods that help us understand economic phenomena.
Twenty years of the Spatial Econometrics Association
Leonidas Doukissas: As Chairman of the Spatial Econometrics Association, how do you assess the Association’s role in the development of the field over the past twenty years?
Giuseppe Arbia: I am very proud of what has happened. Yesterday, at the conference, I presented a graph showing the number of publications in the field from 1974 to the present. After the Association was established in 2006, there was a dramatic increase in publications. At that time, there may have been only 10 to 15 papers per year; today there are more than 1,000. Part of the credit belongs to the Association.
R, Python, and young researchers
Leonidas Doukissas: What advice would you give to young researchers working with spatial data and the broader open-source ecosystem, including R and Python?
Giuseppe Arbia: It is impossible today to imagine research without R and Python. Their importance lies in making it possible to put new techniques and methodologies into practice, as well as in the enormous progress made in computational tools.
Many contemporary methods are highly computationally intensive and could not have been implemented in the past. This applies more generally to empirical-data analysis. I would therefore advise every student and young researcher to become proficient in computing and coding.
The major challenge for the next decade
Leonidas Doukissas: How do you see the future of spatial statistics and spatial econometrics over the next decade?
Giuseppe Arbia: That is a very difficult question. Until now, spatial econometric methods have depended heavily on the specification of tools such as the connectivity matrix. This is a technical instrument that incorporates topology, but its definition is always subjective to some extent.
I believe that the major challenge for the future is to overcome this limitation, so that the decision about “what is close to what” can be specified in an objective way.
Closing
Leonidas Doukissas: Professor Arbia, thank you very much for this conversation. It is a great honour for the Hellenic Spatial Statistics Lab to host your reflections and perspective.
Giuseppe Arbia: The honour is mine. Thank you very much.