A strong positive correlation between the educational attainment of parents and their children is well documented. Determining whether this relationship is due to nature (selection) or nurture (causal factors) is both a challenge and an important policy issue. We use the Vietnam era draft lottery and educational exemptions as a "natural experiment" to address this issue. Substantially more men attended university during this war, creating a cohort of fathers many members of which would not normally have enrolled in tertiary education (TE). Using US Current Population Survey (CPS) and Study of Income Dynamics (PSID) data on the father's and his children's TE involvement, we find that, for this war cohort, the intergenerational transmission leading their children to enrollment in TE is at least as high (CPS) or even higher (PSID) than that of control cohorts. In the context of university attendance in the US, these findings suggest that nurture plays an important additional role.

Guaranteed renewability is a prominent feature in many health and life insurance markets. It is well established in the literature that, when there is (only) risk type uncertainty, the optimal GR contract with renewal price set at the actuarially fair price for low risk types provides full insurance against reclassification risk. We develop a model that includes unpredictable (and unobservable) fluctuations in demand for life insurance as well as changes in risk type (observable) over individuals' lifetimes. The presence of demand type heterogeneity leads to the possibility that optimal GR contracts may have a renewal price that is either above or below the actuarially fair price of the lowest risk type in the population. Individuals whose type turns out to be high risk but low demand renew more of their GR insurance than is efficient due to the attractive renewal price. This results in incomplete insurance against re-classification risk. Although a first best efficient contract is not possible in the presence of demand type heterogeneity, the presence of GR contracts nonetheless improves welfare relative to an environment with only spot markets. Our results also apply to a comparison of environments with short versus long term (front loaded) insurance contracts.

The main contribution of Davies and Hoy (Am Econ Rev 85:980–986, 1995) was a “necessary and sufficient” condition for comparing inequality between income distributions according to the principle of transfer sensitivity (PTS). Chiu (Soc Choice Welf 28:375–399, 2007) showed that although the condition is sufficient, it is not necessary. In this paper, we provide the correct necessary and sufficient condition, and demonstrate with a simple example how the corrected condition allows for more pairs of distributions to be ranked by PTS. The correction clarifies the connection between Lorenz curve comparisons and inequality rankings when the curves intersect.

Smooth handling of pedestrian interactions is a key requirement for Autonomous Vehicles (AV) and Advanced Driver Assistance Systems (ADAS). Such systems call for early and accurate prediction of a pedestrian's crossing/not-crossing behaviour in front of the vehicle. Existing approaches to pedestrian behaviour prediction make use of pedestrian motion, his/her location in a scene and static context variables such as traffic lights, zebra crossings etc. We stress on the necessity of early prediction for smooth operation of such systems. We introduce the influence of vehicle interactions on pedestrian intention for this purpose. In this paper, we show a discernible advance in prediction time aided by the inclusion of such vehicle interaction context. We apply our methods to two different datasets, one in-house collected - NTU dataset and another public real-life benchmark - JAAD dataset. We also propose a generic graphical model Factored Latent-Dynamic Conditional Random Fields (FLDCRF) for single and multi-label sequence prediction as well as joint interaction modeling tasks. FLDCRF outperforms Long Short-Term Memory (LSTM) networks across the datasets ($\sim$100 sequences per dataset) over identical time-series features. While the existing best system predicts pedestrian stopping behaviour with 70\% accuracy 0.38 seconds before the actual events, our system achieves such accuracy at least 0.9 seconds on an average before the actual events across datasets.

Point set registration presents unique significance in Lidar-based intelligent vehicle localization and mapping. It involves registering point sets of the same scene observed from different positions by determining their relative spatial transformation. However, due to the noise and outliers in the point sets and initial misalignment, existing methods suffer from the issues of low accuracy or large computational cost. In this paper, we propose a novel Bayesian state space model to describe the sequential point registration problem. Specifically, we specify the transformations to be the latent states and further assume that they vary smoothly across time. The point clouds are then represented as Gaussian mixture models that change accordingly with the transformation. We then develop a stochastic variational Bayesian inference algorithm to learning the distributions of the transformation, which automatically strike a balance between mapping every two consecutive point clouds and the temporal smoothness of the transformation. Experimental results based simulated data show that the proposed variational Bayesian point set registration (VB-PSR) algorithm achieves higher accuracy with comparable or less time and resources, in comparison with the state- of-the-art methods.

The simultaneous localization and mapping (SLAM) problem involves using the measurements of sensors to construct an environmental map, while simultaneously recovering the vehicle trajectory within this map. There are broadly two strategies for SLAM: on-line and off-line. In this paper, we focus on the off-line SLAM (a.k.a. full SLAM) problem and propose a variational Bayes inference algorithm to address it. Specifically, the intractable posterior distribution of the vehicle poses given the measurements is approximated by a tractable variational distribution, resulting in estimates of the vehicle poses as well as their uncertainties. In contrast with the existing off-line methods, the inverse variances of the additive noise are updated along with the posterior distribution instead of being fixed, thus, the proposed method is robust to unknown noises. Furthermore, the computational complexity of the proposed method is only linear in the number of frames and the computational bottleneck of the algorithm can be easily parallelized to achieve further acceleration. Numerical results show that the proposed method is insensitive to the selection of the noise parameters. More importantly, it is superior in efficiency to the state-of-the-art method, especially for large-scale SLAM problems.

Historically, the measurement of poverty has been based on a set of key axioms but limited to a static framework. Features such as multidimensionality and the time factor have only recently been incorporated into poverty measurement. Several alternative approaches have been proposed. This chapter presents a review of the literature that proposes how to conceptualize or develop specific measures of inter-temporal or lifetime poverty, addressing measurement issues, alternative perspectives on frequency, and concerns about data availability. Reflecting this budding literature, the chapter presents alternative axioms that have been proposed to capture the various approaches to inter-temporal or lifetime poverty. A variety of combinations and alternative formulations of these axioms is considered. Concluding remarks underline the need for further research and focus on the role of empirical analysis in shedding light on issues such as mitigation and chronic poverty.

This chapter addresses ethical and practical challenges for third party sharing of genetic information. Contexts covered range from the usual patient-doctor relationship to researchers (both private and public), employers, insurers, and criminal investigations. Third party interests are identified along with their reasons for wanting or not wanting access to the information available. All scenarios raise critical concerns about privacy and discrimination, as well as potential effectiveness of genetic information and research. There exist wide ranging legal protections in many countries to protect against the sharing of personal genetic information with employers or insurance companies. In areas such as research and the doctor-patient relationship there is still much discussion necessary to deal with how to trade off the rights of individuals to privacy and the rights of interested third parties. The complexity of the issues involved and the rapid progress in the field of genomics create significant challenges.

In this paper, we study the actor-action semantic segmentation problem, which requires joint labeling of both actor and action categories in video frames. One major challenge for this task is that when an actor performs an action, different body parts of the actor provide different types of cues for the action category and may receive inconsistent action labeling when they are labeled independently. To address this issue, we propose an end-to-end region-based actor-action segmentation approach which relies on region masks from an instance segmentation algorithm. Our main novelty is to avoid labeling pixels in a region mask independently - instead we assign a single action label to these pixels to achieve consistent action labeling. When a pixel belongs to multiple region masks, max pooling is applied to resolve labeling conflicts. Our approach uses a two-stream network as the front-end (which learns features capturing both appearance and motion information), and uses two region-based segmentation networks as the back-end (which takes the fused features from the two-stream network as the input and predicts actor-action labeling). Experiments on the A2D dataset demonstrate that both the region-based segmentation strategy and the fused features from the two-stream network contribute to the performance improvements. The proposed approach outperforms the state-of-the-art results by more than 8% in mean class accuracy, and more than 5% in mean class IOU, which validates its effectiveness.

Guaranteed renewability is a prominent feature in many health and life insurance markets. It is well established in the literature that, when there is (only) risk type uncertainty, the optimal GR contract with renewal price set at the actuarially fair price for low risk types provides full insurance against reclassification risk. We develop a model that includes unpredictable (and unobservable) fluctuations in demand for life insurance as well as changes in risk type (observable) over individuals' lifetimes. The presence of demand type heterogeneity leads to the possibility that optimal GR contracts may have a renewal price that is either above or below the actuarially fair price of the lowest risk type in the population. Individuals whose type turns out to be high risk but low demand renew more of their GR insurance than is efficient due to the attractive renewal price. This results in incomplete insurance against re-classification risk. Although a first best efficient contract is not possible in the presence of demand type heterogeneity, the presence of GR contracts nonetheless improves welfare relative to an environment with only spot markets. Our results also apply to a comparison of environments with short versus long term (front loaded) insurance contracts. JEL-Codes: D800, D860, G220.

Loss Coverage: Why Insurance Works Better with Some Adverse Selection, Edited by R. Guy Thomas, Cambridge: Cambridge University Press, 2018, 274pp, ISBN: 978-1-107-49590-6 – ERRATUM - Volume 13 Issue 2

We propose a new deep learning based system for short term prediction of pedestrian behavior in front of a vehicle. To achieve this, we first develop a framework for class specific object tracking and short term path prediction based on a variant of a Variational Recurrent Neural Network (VRNN), which incorporates latent variables corresponding to a dynamic state space model. The low level visual features learned from this system were found to be highly informative for the discrete intention prediction task (i.e., predicting whether a pedestrian is stopping or crossing), and achieved high performance on the Daimler benchmark. This is despite a much smaller training dataset than is normally used for training deep learning models. To the best of our knowledge, we are the first to apply deep learning to this problem without using externally trained pedestrian pose estimation systems. Our system performs comparable to the state-of-the-art approach that relies on pose estimation, and runs in real time.

Lidar segmentation is a common preprocessing step for performing ground object detection in autonomous vehicle applications. However, many common approaches are affected by at least one of the following three issues: they are prone to under-segmentation/over-segmentation; they are not able to effectively handle missing points caused by black objects or bright sunlight; or they are too complex for real time use on commodity hardware. In this paper we propose a modification to distance based segmentation algorithms, which is able to improve the performance without incurring significant computational cost. Numerical evaluations on the KITTI dataset confirm the methods applicability.

Autonomous Vehicles (AV) and Advanced Driver Assistance Systems (ADAS) require smart built-in systems to recognize pedestrian intention to avoid casualties. Recent approaches towards predicting pedestrian intention use motion information and his/her situational awareness to distinguish between intention of either crossing or stopping before a vehicle. We investigate the power of including contextual information into this problem. Examples of contextual cues include pedestrian position with respect to the curb of a road and relative locations of the vehicles with respect to the pedestrian in the scene. We propose a graphical model Factored Latent Dynamic Conditional Random Fields (FLDCRF) to integrate contextual and motion features of a pedestrian as well as to capture the intrinsic relationship between intention and action of a pedestrian. We show that the inclusion of context information aids early prediction of a pedestrian intention and the motion information provides more accuracy to the predicted output, resulting in an early and accurate joint model to predict pedestrian intention. We also show that FLDCRF performs better over the concatenated motion and context features when compared with existing state-of-the-art machine learning algorithms. The results show accurate estimates of pedestrian intention with high average probabilities (>0.7) as early as 1.2 second before the performed action.

Many approaches for tracking objects in lidar data have been proposed in recent years. However, most practical real time systems assume that clean segmentation of lidar points into individual objects can be achieved. Unfortunately, efficient lidar segmentation approaches are prone to under-segmentation when objects are very close to each other; one solution is to introduce additional segmentation steps into the tracking process. In this paper we propose a new method to address this task with distance dependent Chinese Restaurant Processes (dd-CRP) equipped with a shape prior defining possible object shapes. By adding constraints to the segmentation model, we are able to further improve stability of segmentation and tracking. Experiments on real datasets show the advantage of this approach over a baseline object tracking pipeline.

This note develops a new approach to measuring discrimination. A partial ordering of discrimination patterns is proposed that is consistent with the properties of second-degree stochastic dominance (SSD), which are related to changes in the distributions of either the reference (advantaged) or comparison (disadvantaged) group, while keeping the other group's distribution unchanged. Furthermore, a corresponding summary index is derived. This index provides a complete ordering to rank discrimination patterns and also satisfies the principles of SSD.

The simultaneous localization and mapping (SLAM) problem is concerned with using sensor data to build an environmental map, while also localizing an autonomous agent within this map. Two approaches are currently prevalent (Bayesian filtering and graph-based optimization), however these both involve approximations and have the potential to be improved. In this paper, we propose novel high-performance SLAM algorithms derived from variational Bayes inference. By using mean-field type approximations, the resulting computational complexity is linear. We also add an empirical Bayes assumption to improve the flexibility of the inference. Experiments are conducted on both synthetic data and real RGB-D images. The proposed approach achieves 42% average error reduction in all scenarios on synthetic data, and 26% average error reduction on real images (with respect to two other baseline algorithms).

Semantic segmentation is an important task for autonomous vehicle navigation in off road environments. However, several natural factors make this problem uniquely challenging. For example, road segmentation is often difficult under heavy shadow or steel terrain, and dangerous muddy water puddles may have the similar visual appearance to dirt road surfaces (and thus are hard to identify). To tacule these challenges, we present a semantic segmentation system based on a two-stage hierarchical fusion pipeline. The first stage improves the road segmentation by effectively fusing information from camera and 3D Lidar point cloud. The second stage is dedicated to detecting water puddles, based on the results from the first stage. Due to the parallelized architecture, our system can be deployed for real-time applications. We achieved an F1 score of around 93% for road segmentation and 80% for water puddle segmentation at more than 10 Hz.