Software development is a collaborative effort that leverages existing code, which reduces the cost of developing new software. That said, leveraging existing code exposes coders to vulnerabilities, because existing code might contain programming bugs. We study the formation of dependency networks among software repositories, guided by a strategic model of network formation with observable and unobservable heterogeneity. We estimate costs, benefits, and link externalities of a network of 696, 790 directed dependencies between 35, 473 repositories of the Rust programming language using a scalable statistical algorithm. We find evidence of a positive externality exerted on other coders when coders create dependencies. What is more, we show that coders are likely to link to more popular packages of the same and of other software types. We adopt models for the spread of infectious diseases to measure a package's systemicness as the number of downstream packages a vulnerability would affect. Systemicness is highly skewed with the most systemic repository affecting almost 90% of all repositories only two steps away. Finally, we show that protecting only the ten most important repositories reduces vulnerability contagion by nearly 40%.

Neural network representations of simple models, such as linear regression, are being studied increasingly to better understand the underlying principles of deep learning algorithms. However, neural representations of distributional regression models, such as the Cox model, have received little attention so far. We close this gap by proposing a framework for distributional regression using inverse flow transformations (DRIFT), which includes neural representations of the aforementioned models. We empirically demonstrate that the neural representations of models in DRIFT can serve as a substitute for their classical statistical counterparts in several applications involving continuous, ordered, time-series, and survival outcomes. We confirm that models in DRIFT empirically match the performance of several statistical methods in terms of estimation of partial effects, prediction, and aleatoric uncertainty quantification. DRIFT covers both interpretable statistical models and flexible neural networks opening up new avenues in both statistical modeling and deep learning.