Research
SMaC: Spatial Matrix Completion Method
G.Grossi, A.Mattei, G.Papadogeorgou
Synthetic control methods are commonly used in panel data settings to evaluate the effect of an intervention. In many of these cases, the treated and control time series correspond to spatial areas such as regions or neighborhoods. We work in the setting where a treatment is applied at a given location and its effect can emanate across space. An area of certain size around the intervention point is assumed to be the treated area. Synthetic control methods can be used to evaluate the effect that the treatment had in the specific area, but it is often unclear how far the treatment's effect propagates. Therefore, researchers might consider treated areas of different sizes and apply synthetic control methods separately for each area. However, this approach ignores the spatial structure of the data, and can lead to efficiency loss in spatial settings. We propose to deal with these issues by developing a Bayesian spatial matrix completion framework that allows us to predict the missing potential outcomes in the different areas around the intervention point while accounting for the spatial structure of the data. Specifically, the missing time series in the absence of treatment for the treated areas are imputed using a weighted average of control time series, where the weights are assumed to vary smoothly over space according to a Gaussian process. Our motivating application is the construction of the first line of the Florentine tramway, which could have had an effect on the prevalence of businesses in the neighbourhood of the construction site, and at various distances from the tramway stops.
AutoSynth: using autoencoders to build synthetic indicator for socio-economic development
G.Grossi, E. Rocco
Synthetic indicators for socio-economic development has the main goal of summarizing a wide set of information into a simple indicator, that in principle should represent the underlying phenomenon. In fact, several decisions could affect the final results of the indicator, leaving space for improvements and more honest feature representation. In this work, we propose a novel use for neural networks to build socioeconomic indicators, encoding a possible large information set, within single or multiple synthetic indexes, we call this proposal AutoSynth. In particular, we encode such information using an autoencoder, a neural network method to represent in a lower dimensionality space a matrix of features. We apply such a method to the evaluation of socio-economic developments of suburban areas in Florence, and we test the performance of our model against some golden standard methods using a stress test.
The policy is always greener: impact heterogeneity of Covid-19 vaccination lotteries in the US
G.Grossi
Covid-19 vaccination has posed crucial challenges to policymakers and health administrations worldwide. In addition to the pressure posed by the pandemic, government administration has to strive against vaccine hesitancy, which seems to be considerably higher concerning previous vaccination rollouts. To increase the vaccination protection of the population, Ohio announced a monetary incentive as a lottery for those who decided to vaccinate. This first example was followed by 18 other states, with varying results. In this paper, we want to evaluate the effect of such policies within the potential outcome framework, using the penalized synthetic control method. We treat with a panel dataset and estimate causal effects at a disaggregated level in the context of staggered treatment adoption. We focused on policy outcomes at the county, state, and supra-state levels, highlighting differences between counties with different social characteristics and time frames for policy introduction. We also studied the nature of the treatment effect to see whether the impact of these monetary incentives was permanent or only temporary, accelerating the vaccination of citizens who would have been vaccinated in any case.
Synthetic control method in presence of Interference: the direct and spillover effect of light rail on neighborhood retail activities
G.Grossi, P.Lattarulo, M.Mariani, A.Mattei, O.Oner
In recent years, Synthetic Control Group (SCG) methods have received great attention from scholars and have been subject to extensions and comparisons with alternative approaches for program evaluation. However, the existing methodological literature mainly relies on the assumption of non-interference. We investigate the use of the SCG method in panel comparative case studies where interference between the treated and the untreated units is plausible. We frame our discussion in the potential outcomes approach. Under a partial interference assumption, we formally define relevant direct and spillover effects. We also consider the "unrealized" spillover effect on the treated unit in the hypothetical scenario that another unit in the treated unit's neighborhood had been assigned to the intervention. Then we investigate the assumptions under which we can identify and estimate the causal effects of interest, and show how they can be estimated using the SCG method. We apply our approach to the analysis of an observational study, where the focus is on assessing direct and spillover causal effects of a new light rail line recently built in Florence (Italy) on the retail density of the street where it was built and of the streets in the treated street's neighborhood.
Bayesian Principal Stratification with longitudinal data and truncation by death
G.Grossi,, M.Mariani, A.Mattei, F.Mealli
Self-enterprise programs are used to promote youth and female employment while stimulating the formation of new businesses and jobs. In this study we are going to analyze the results of the Tuscany region's "Fare Impresa" program, facing with treatment evaluation in presence of censored outcomes. In many causal studies, outcomes are 'censored by death,' in the sense that they are neither observed nor defined for units who die. In order to estimate causal quantities in these contexts, Frangakis and Rubin, 2002 propose the principal stratification approach, which defines partitions of the units in latent groups (the principal strata). In such studies, focus is usually on the stratum of 'always survivors' up to a single fixed time s. Building on a recent strand of the literature, we propose an extended framework for the analysis of longitudinal studies, where units can die at different time points, and the main endpoints are observed and well-defined only up to the death time. We develop a Bayesian longitudinal principal stratification framework, where units are cross-classified according to the longitudinal death status. Under this framework, focus is on causal effects for the principal strata of units that would be alive up to a time point s irrespective of their treatment assignment, where these strata may vary as a function of s. We can get precious insights on the effects of treatment by inspecting the distribution of baseline characteristics within each longitudinal principal stratum, and by investigating the time trend of both principal stratum membership and survivor-average causal effects.
Fighting Tax Evasion with intergovernemental cooperation: direct and indirect effects
A.Angeli, G.Grossi, P.Lattarulo, M.G.Pazienza
Our study focuses on a cooperation policy between the Revenue Agency and municipalities, devoted to tackling tax evasion. Statements proposed by many international public advisors and scholars focus on the collaboration among institutions as a strategy to contrast evasion. Indeed, institutional collaborative governance is an acknowledged approach in public administration to pursue collective goals in an interdependent multilevel framework. Nonetheless, only little empirical evidence is provided by scholars on the effects of this policy, either on the recovered taxes or on the tax compliance.
A cooperation policy between Italian Revenue Agency and municipalities to detect evasion of central taxes started in Italy in 2009 and is currently ongoing. We found that the Agency’s successful auditing effort can be an incentive for local administrations to take part in the reporting activity, just as a lack of audits by the tax offices can be a deterrent. Moreover, the growing experience in detecting national taxes that characterize those municipalities participating to the policy for consecutive years results in more qualified reporting activity and growing tax recovery. This represents a case of successful “learning by doing” activity. Our findings suggest that increasing experience in detecting national taxes extends to other municipal revenues and continues in the subsequent years. Policies to tackle tax evasion do not only lead to an immediate tax recovery but induce higher tax compliance that remains in the following years. This implies a most needed benefit for the community, as in fact, it is widely shared among scholars the key role of wide and long-term changes in the individual and institutional attitude toward tax evasion