@article{CaliendoKuennWeissenberger2020,
  author    = {Caliendo, Marco and K{\"u}nn, Steffen and Weissenberger, Martin},
  title     = {Catching up or lagging behind?},
  series = {Research policy : policy, management and economic studies of science, technology and innovation},
  volume    = {49},
  journal   = {Research policy : policy, management and economic studies of science, technology and innovation},
  number    = {10},
  publisher = {Elsevier},
  address   = {Amsterdam},
  issn      = {0048-7333},
  doi       = {10.1016/j.respol.2020.104053},
  pages     = {14},
  year      = {2020},
  abstract  = {From an active labor market policy perspective, start-up subsidies for unemployed individuals are very effective in improving long-term labor market outcomes for participants. From a business perspective, however, the assessment of these public programs is less clear since they might attract individuals with low entrepreneurial abilities and produce businesses with low survival rates and little contribution to job creation, economic growth, and innovation. In this paper, we use a rich data set to compare participants of a German start-up subsidy program for unemployed individuals to a group of regular founders who started from non-unemployment and did not receive the subsidy. The data allows us to analyze their business performance up until 40 months after business formation. We find that formerly subsidized founders lag behind not only in survival and job creation, but especially also in innovation activities. The gaps in these business outcomes are relatively constant or even widening over time. Hence, we do not see any indication of catching up in the longer run. While the gap in survival can be entirely explained by initial differences in observable start-up characteristics, the gap in business development remains and seems to be the result of restricted access to capital as well as differential business strategies and dynamics. Considering these conflicting results for the assessment of the subsidy program from an ALMP and business perspective, policy makers need to carefully weigh the costs and benefits of such a strategy to find the right policy mix.},
  language  = {en}
}
@article{ThamsenBeilharzVinhThuyTranetal.2020,
  author    = {Thamsen, Lauritz and Beilharz, Jossekin Jakob and Vinh Thuy Tran, and Nedelkoski, Sasho and Kao, Odej},
  title     = {Mary, Hugo, and Hugo*},
  series = {Concurrency and computation : practice \& experience},
  volume    = {33},
  journal   = {Concurrency and computation : practice \& experience},
  number    = {18},
  publisher = {Wiley},
  address   = {Hoboken},
  issn      = {1532-0626},
  doi       = {10.1002/cpe.5823},
  pages     = {12},
  year      = {2020},
  abstract  = {Distributed data-parallel processing systems like MapReduce, Spark, and Flink are popular for analyzing large datasets using cluster resources. Resource management systems like YARN or Mesos in turn allow multiple data-parallel processing jobs to share cluster resources in temporary containers. Often, the containers do not isolate resource usage to achieve high degrees of overall resource utilization despite overprovisioning and the often fluctuating utilization of specific jobs. However, some combinations of jobs utilize resources better and interfere less with each other when running on the same shared nodes than others. This article presents an approach for improving the resource utilization and job throughput when scheduling recurring distributed data-parallel processing jobs in shared clusters. The approach is based on reinforcement learning and a measure of co-location goodness to have cluster schedulers learn over time which jobs are best executed together on shared resources. We evaluated this approach over the last years with three prototype schedulers that build on each other: Mary, Hugo, and Hugo*. For the evaluation we used exemplary Flink and Spark jobs from different application domains and clusters of commodity nodes managed by YARN. The results of these experiments show that our approach can increase resource utilization and job throughput significantly.},
  language  = {en}
}