Gianluca Manzo's Empirically-oriented Work
What micro- and network-level mechanisms explain differentiation, change and stability at the macro-level? This is the main, general question that I address in my empirically-oriented work. In particular, I am interested in the way social interactions work at the same time as amplifiers of individual-level differences and as counterbalancing forces limiting macro-level differentiation and change. Educational inequalities, relative deprivation, status hierarchies, and, more recently, innovation diffusion and migration dynamics constitute the empirical phenomena that I study from this perspective.
// Migration Dynamics (current project)
This is my newest line of research conducted in collaboration with Filiz Garip (Harvard University). The project, which is in its enfancy, aims to establish a bridge between her deep, empirically-informed knowledge of Mexico-US migratory fluxes and my work on formal modeling of relative deprivation.
Our objective is to build a data-driven agent-based computational model that explains migratory dynamics as a self-regulated process in which actors’ choices to migrate dynamically change the social networks they are embedded in thus implicitly affecting their own and other potentials migrant’s satisfaction threshold and reference points. Our hypothesis is that, since migration choices partly depend on actors’ network-based reference points, the endogenously generated network changes unintentionally work as a migration stabilizer.
// Diffusion of Innovations (current project)
This is a new line of research conducted in collaboration with anthropologists, archeologists, and psychologists. Our common interest is to understand whether cultural and status-related factors help to explain spatial and temporal patterns of diffusions of technical innovations (in particular, in the field of pottery) in different areas of the world, namely India, Cameroun, Kenya, Ethiopia, Philippines, and Ecuador.
In this large, multi-partner, and multi-disciplinary project, I have the responsibility of the modeling package. More particularly, on the one hand, I am working to complement the qualitative, in-depth interview-based data collected by the anthropologists and psychologists with more standardized information on individual’ attitudes toward innovation and on the advice networks in which individuals (namely, potters) are embedded. On the other hand, I am asked to use this qualitative and quantitative information as basis to develop a data-driven agent-based computational model that formalizes hypotheses on the micro-level and network-related mechanisms underlying the observed diffusion dynamics.
This project, named DIFFCERAM for DIFFusion of CERAMic Techniques and Styles, was granted (379000 Euros) by the French Agency for Research (ANR) in 2012. The French ethno-archeologist Valentine Roux (CNRS, Préhistoire & Technologie) is the project's PI, and Simone Gabbriellini (CNRS, GEMASS) is my collaborator on the modeling package.
// Status Hierarchies (current project)
This is a line of research started in 2010 and is conducted in collaboration with Delia Baldassari (New York University). It is part of my interest for the mechanisms underlying the subjective and symbolic dimensions of social stratification (see below my work on relative deprivation).
In this project, we focus on the micro-level dynamics of deference exchange that can explain some of the macro-level features regularly observed in real-world status hierarchies, namely 1/ the gap between actors’ quality and the amount of deference they are able to secure; 2/ the temporally varying amount of rank switches; and 3/ the high level of distributive inequality. To address these questions, we started with studying, and attempting to replicate, previous formal models of status and reputation dynamics. Then, we develop our own model with the aim to increase the realism of the micro-level mechanisms driving deferential gesture (heuristics versus utility maximization). In addition, we introduced more realistic network constraints defining who can exchange deference with whom.
The model was studied by means of agent-based simulations. Our main result is that the combined action of mechanisms supposed to trigger status asymmetry with mechanisms supposed to contain it in fact crucially depends on the amount of individual-level heterogeneity postulated as well as on actors’ propensity to interact with status-dissimilar others.
Learn more on my publications and talks on "Status Hierarchies".
// Relative Deprivation and Happiness
I started this line of research in 2008 as an attempt to generalize my analysis of social stratification to its subjective dimension. In particular, I moved from the study of the effects of actors’ social belonging on their objective social opportunities (see below my work on educational opportunities), to the analysis of actors' perception of these opportunities.
With this aim, I started by re-implementing and re-analyzing early game-theoretic formal models of relative deprivation within agent-based computational framework. These models provide indeed crucial insights on the conditions under which more opportunity can in fact lead to smaller proportion of satisfied actors. Then, I extended these models in two ways: 1/ I introduced a set of mechanisms that are assumed to drive the intensity of actors’ feelings of satisfaction; 2/ building on structural analyses of relative deprivation, I introduced dyadic ties among artificial agents in order to represent actors performing neighborhood-based comparisons rather than population-based comparisons.
Extensive agent-based simulations of these models led to two main results: 1/ when it is postulated that actors’ envy inversely depends on the proportion of actors “in the same boat” (according to Merton’s original intuition), the sought-after pattern “ more opportunities, more satisfied actors, less intense individual-level feeling of dissatisfaction” is rarely observed; 2/ the more the network connecting agents is sparse and contains low-degree nodes, the more frequent the pattern “more opportunities, lower dissatisfaction levels”.
The results of this formal, theoretical analysis have informed a regression-based analysis of subjectively-reported levels of satisfaction in France based on a large-scale survey conducted in 2009 by the GEMASS. In particular, in this classic, meaning statistics-based, analysis, I assess the extent to which the association between actors’ income and subjectively-reported satisfaction disappear when several indicators of the social comparisons operated by actors are introduced into the picture, thus suggesting complex interaction between objective and subjective factors.
Macroscopic patterns and trends in educational inequalities represent my oldest research topic. I have been working on it since my bachelor dissertation (2000-2001) in which I explored the possibility to use artificial neural networks as a descriptive tool to predict individuals’ educational outcomes in Italy during the twentieth century.
In my doctorate dissertation (2002-2006) in which I compared patterns of, and temporal trends in, educational inequality in France and in Italy, I used advanced multivariate statistics for categorical data (namely, log-linear and log-multiplicative models). It is at that time that I started to regard simulation-based formal models (namely, agent-based computational models) as complementary to multivariate statistics.
My work in this field thus employs, on the one hand, standard quantitative methods to analyze large-scale survey data in order to reveal the macroscopic configuration and evolution of educational stratification among socio-economic groups, and, on the other hand, agent-based computational models to test hypotheses about the social mechanisms underlying these macro-level regularities.
By using this combination of methods, I demonstrated that the observed level of educational inequality can only be reproduced if actors’ heterogeneity in terms of ability and benefit/costs perceptions is coupled with strongly segregated friendship networks.
In my view, this line of research has produced four major (set of) publications. The first reported on results of my early descriptive analysis of Italian data performed by means of artificial neural networks and was published (in Italian) in 2003 in Sociologia e Ricerca Sociale. The second included a topological multi-matrix log-linear model of the cross-sectional structure of educational stratification in France and was published in Quality and Quantity in 2006.
The third is a book that summarizes and develops my PhD dissertation. It presents a comparative analysis of French and Italian educational stratification and is based on a combination of multivariate statistics and agent-based computational modeling which was published in 2009 by the Presses de l’Université Paris-Sorbonne.
The fourth is an article that extends the agent-based computational model of educational choices at the heart of the 2009 book, applies it to new French large-scale survey data, and refines the model’s analysis and comparison with empirical data. This article was published in 2013 in Comparative Social Research. To me, this article provides the most solid defense of my thesis according to which amplifying effects fueled by network-related externalities are needed to generate the amount of educational inequalities usually observed at the macro-level in contemporary societies.
Learn more on my publications and talks on "Educational Inequalities".