Denise Li (p), André Chagas

This paper studies neighborhood effects in consumption behavior using consumption survey data of households from the city of São Paulo, Brazil. The data has disaggregated information about the expenses of 3,120 households surveyed between 2008 and 2013. We also have access to the exact address of all families, which helped us to define more precisely household's peer group based on geographical proximity. To address endogeneity problems typical of peer effect studies due to the simultaneous presence of contextual and correlated effects, we use an adapted spatial autoregressive model. However, this type of model imposes specific network structure to the data (such as observation of intransitive triads, i.e., “neighbors of neighbors who is not neighbors themselves”). As an alternative identification strategy, we use an instrumental variable fixed effects approach. An ideal instrument for peer’s consumption expenditure is exogenous variation in idiosyncratic expenditure shocks faced by peers. To overcome our lack of data about these shocks, we propose the use of an instrument created by exploiting variation in the excess demand for public childcare across age cohorts within São Paulo’s neighborhoods. The same instrument was used to find a positive effect of childcare use on maternal employment. We consider that the probability of obtaining a public daycare slot affects a household's consumption expenditure but does not directly affect their peers' consumption expenditure. The relevance of the research question for the correct evaluation of consumption intervention programs or of other policies that can have indirect spillover effects motivated many recent studies. Mainly due to data restriction, there still is space in the literature for contribution, which depends on the availability of new data and creative identification strategies to overcome the endogeneity problems. This study is unique for the Brazilian case, and the results contribute to the understanding of consumption network effects in less developed countries.

Joao Carlos Claudio, Katja Heinisch (p), Oliver Holtemöller

This paper presents a mixed-data sampling (MIDAS) model to nowcast regional real GDP growth based on a newly constructed quarterly series measuring East German economic activity. More specifically, our model takes into account past quarterly East German real GDP growth, contemporaneous and past quarterly German real GDP growth and various monthly business cycle indicators available for East Germany. Our approach also includes a forecast for current quarter German real GDP growth. By applying this setting, individual MIDAS models are able to outperform conventional autoregressive (AR) and several autoregressive distributed lag (ARDL) models. Evidence based on single indicator models suggests that survey data on the situation in the construction sector and the expectations in the wholesale trade sector are among the most useful to monitor short-term economic activity in East Germany. Averaging forecasts from the MIDAS models further improve our results.

Ever Enrique Castillo Osorio (p), Hwan-Hee Yoo

After the occurrence of a disaster in urban areas, many people can be trapped inside buildings such as houses, offices, etc. In case of this vulnerable people, they need support for a quick evacuation, especially in large new buildings that follow design trends too complex, and that at the time of panic could turn into a maze when people try to find the emergency exits nearest.
Nowadays, current design technologies such as Building Information Modeling (BIM) let us to know the functional distribution of the geometric context inside buildings. Data can be taken of the BIM for evacuation activities of people indoor buildings towards safe places. However, they must be integrated efficiently, in order to be feasible for such activities. On the other hand, the study of algorithms for generation of evacuation routes has had notable advances in the scientific field, currently applying criteria based on machine learning and artificial intelligence. These criteria allow algorithms to be more efficient under conditions of time and movement of people to secure areas.
This work aims to develop a method for the calculation of shorter evacuation routes inside buildings. The first step is to obtain the simplified geometry of the building taken from the BIM. The geometry includes walls, doors, corridors, ramps and others facilities. As a second step, the information of the geometry with its basic attribute table is transferred in a spatial database. Attribute tables are processed under the conception of Geographic Information System (GIS), that is, each table stores multiple characteristics of the elements. In the third step is applied the multicriteria analysis in the geometry. It considers the allocation of weights based on information of the design and structural context of the building. Finally, the fourth step is the calculation of the most feasible evacuation route under the concept of machine learning. This is calculated through an algorithm generated as a variant of the A* algorithm, and the use of the Dijkstra's algorithm as additional criteria for obtaining the shortest route among various feasible routes.
The experimental results show the importance of the allocation of weights in the context of multicriteria analysis and the use of the algorithm that allows determining the safest evacuation routes for vulnerable people in search and rescue activities.

Irfani Fithria Ummul Muzayanah (p), Djoni Hartono (p), Suahasil Nazara, B Raksaka Mahi

The determination of social capital investment is not only affected by individual’s socio – economic characteristics, but also shaped by the characteristics of physical environment. High density urban form also claimed as suitable urban form to enhance social capital. However, the relationship between these two has not been empirically well – explored and competing results emerge in the literatures. Accordingly, this study aims to investigate the role of urban physical arrangement or urban form in the social capital formation using metropolitan cities in Indonesia as case study. This study is claimed as first empirical study to investigate the association between urban form and social capital in Indonesia. Multilevel logistic regression is employed to investigate the association between urban form and four dimensions of social capital : trust, reciprocity, social participation and civic engagement. In addition, this study also distinguishes bonding, bridging and linking as three basic forms of social capital. This study reveals that density have no association with bonding, bridging, linking social capital as well as social participation and political participation. While, density is only positively associated with membership and reciprocity. Furthermore, other urban form elements such as street connectivity and the presence of local destination showed negative association with several dimensions of social capital.