SEMINARIOS Y WORKSHOPS

Fecha: 12/11/2020, 11:00

Modalidad: Virtual

Presentador: Alvaro Riascos (Universidad de los Andes)

Autor: Riascos , Alvaro

In this paper we address the accuracy vs. fairness trade-off in state of the art machine learning algorithms for crime prediction. We propose a novel conditional GANs architecture for crime (robberies) prediction in Bogota, capital city of Colombia. The model uses several layers of ConvLSTM neural nets in both the generative and the discriminatory networks. We further condition on past crime intensity maps, weekdays, and holidays. The trained network is able to capture spatio-temporal patterns and outperforms state-of-the-art predictive models such as spatio-temporal Poisson point process, as well as other models trained with the same dataset. Model’s accuracy reaches an area under the Hit Rate – Percentage Area Covered by Hotspots curve of 0.86. However, our predictions suggest that there is a potential bias with heterogeneous effects on vulnerable populations. We address the fairness consequence of this model in low income vs. high income residents by estimating a calibration test conditional to these protected variables. Finally, we introduce a fairness – accuracy balancing technique that quantifies the trade-offs between accuracy and fairness in this type of models. This technique notably reduces bias with a marginal effect on accuracy. The previous model addresses the accuracy vs. fairness trade-off when the protected variable is low income vs. high income populations. However, this distinction is quite artificial in our data set since we use a proxy of income based on administrative data records. Therefore, we consider a data driven urban characterization based on city images that better reflect the living, socioeconomic and environmental conditions of the city. To do this we use state-of-the-art crime prediction models of spatio-temporal clustering including environmental covariates extracted through convolutional networks from street-level images. Our model using image features as covariates outperforms a standard model of self-exciting point process and it’s well calibrated, even before we try to fix it using ex post Platt scaling.

Más información

Fecha: 19/11/2020, 11:00

Modalidad: Virtual

Presentador: Anna Bindler (U of Cologne)

Autor: Bindler, Anna

This paper evaluates the effect on crime of creating a fundamental modern-day institution: centralized professional police forces tasked with preventing crime. We study the 1829 formation of the London Metropolitan Police – the first professional force worldwide. Using newly digitized and geocoded crime and police data together with difference-in-differences and pre-post designs, we find evidence of a significant reduction in violent crimes (despite the possibility of off-setting increases in clearance and reporting rates). In contrast, a reduction in property crime is not visible.

Más información

Fecha: 26/11/2020, 11:00

Modalidad: Virtual

Presentador: Luis Baldomero-Quintana (William & Mary)

Autor: Baldomero-Quintana, Luis

This paper provides evidence that domestic trade costs are a source of national comparative advantage. I build a model of international trade and internal geography that considers both international shipping routes and input-output linkages. I use the model to simulate how a large infrastructure project, Ruta del Sol, affects the specialization of Colombia, a country whose exports are highly concentrated in the mining sector. This road improves access to global markets for both mining and manufacturing regions. To quantify the model and estimate its parameters, I use customs administrative records, a transportation survey, and geospatial data generated from both physical and digital road maps. My results indicate that the road project shifts the comparative advantage of Colombia away from mining, and towards manufacturing. Lastly, I show that the change in comparative advantage is larger when industry linkages are considered because access to tradable intermediate inputs benefits the manufacturing sector more than the mining sector. The results demonstrate that national comparative advantage is shaped materially by domestic trade costs, beyond the elements typically analyzed in the literature such as factor endowments, technology and institutions.

Más información

Fecha: 18/03/2021, 11:00

Modalidad: Virtual

Presentador: Nelson Ruiz (URosario/Oxford)

Autor: Ruiz, Nelson

Co-autor:  Miguel Rueda (Emory)

What motivates citizens to donate to a political campaign? The literature has characterized donors as those who wish to receive a direct benefit from an elected official or as citizens who just want to express their affinity for a candidate through a donation. We argue that, in a context of weak parties and term limits, donation patterns over time should differ for both types of donors. Using a close election regression discontinuity design in a sample of Colombian municipalities, we find that donating to the winner of the mayoral race significantly reduces the probability of donating in the next election. Consistent with our theory, such an effect is larger for family members---a group of donors with strong expressive motivations---than for private donors. Moreover, unlike family member donors, private donors are more likely to receive public contracts when donating to an election winner and more likely  to donate in the next election after receiving a public contract.

Más información

Fecha: 25/03/2021, 11:00

Modalidad: Virtual

Presentador: Francesco Bogliacino

Autor: Bogliacino, Francesco

Co-autor: Christian Posso y Juanita Villaveces

We estimate the causal effect of the Colombian land restitution program on access to microcredit for agriculture. We use the timing of the restitution as the source of identification in an event study approach. Using administrative data from the program and all formal credit transactions, we show a significant increase in access to agricultural microcredit. The effects are stronger two years after the restitution when individuals acquire full property rights. Impacts are mainly driven by loans through the Agrarian Bank and supported by guarantees from the Agricultural Guarantee Fund, which are institutions specifically designed to promote investment in agriculture.

Más información

Fecha: 08/04/2021, 11:00

Modalidad: Virtual

Presentador: Samuel Lordemus

Autor: Lordemus, Samuel

Co-autor:  
Noemi Kreif -Rodrigo Moreno-Serra

How do government counterinsurgency efforts affect local public financing during civil conflicts? We investigate this question in the context of the protracted conflict in Colombia. Using data on antinarcotics operations and health transfers from the central government to municipal governments, we employ both panel estimations and instrumental variables to address concerns of endogeneity and omitted variables. We find no clear evidence that counterinsurgency operations causally affect health transfers to municipalities. However, we find indicative evidence that counterinsurgency operations undertaken in a given year increase the intensity of non-state violence in the following year. Our findings suggest that armed counterinsurgency interventions by the State should be accompanied by renewed measures to support public healthcare financing for the affected local populations; otherwise, such interventions risk exacerbating the negative consequences of conflict exposure on population health.

Más información

Fecha: 15/04/2021, 11:00

Modalidad: Virtual

Presentador: Manuel Barron (Universidad del Pacifico)

Autor: Barron, Manuel

This paper studies how the high school's identity affects students' sorting in the higher education market in Peru, a country without standardized tests at the end of high school. We first examine whether providing low-income students with mechanisms to signal their ability improves their higher education outcomes. Our results show that graduating from an elite public high school increases higher-education enrollment and quality even without noticeable gains in human capital among marginally admitted students. Likewise, obtaining the IB diploma by the minimum margin increases enrollment and admissions at top private institutions. These effects are more pronounced on a special admissions mode, under which  universities waive admission tests to selected applicants. We also estimate value-added models on higher education outcomes and find consistent evidence with top private institutions statistically discriminating based on sending high schools.

Más información

Fecha: 29/04/2021, 11:00

Modalidad: Virtual

Presentador: Alex Solis

Autor: Solis, Alex

This paper estimates the effects of university loans and grants availability on educational and labor market outcomes, examining the eligibility cutoffs in the Chilean financial aid system that enable a regression discontinuity design. Results show that access to university loans increases degree completion among women, and creates negative effects on labor experience and earnings that revert over time. Also, loans allow students to choose higher-quality programs in fields of study with higher earning prospects. In the long run, women experience positive and significant earnings impact. In contrast, equivalent access to grants does not affect degree completion or labor market outcomes to the extent observed for loans. The limited sensitivity to a reduction in tuition costs observed at the grant cutoff, and the higher effects among students from poorer backgrounds suggest that the effects are driven by credit constraints.

Más información