Chair Agricultural Production and Natural Resource Economics, Technical University of Munich, Munich, Bavaria, Germany.
PLoS One. 2021 Mar 12;16(3):e0247437. doi: 10.1371/journal.pone.0247437. eCollection 2021.
This article investigates the technical efficiency in German higher education while accounting for possible heterogeneity in the production technology. We investigate whether a latent class model would identify the different sub-disciplines of life sciences in a sample of biology and agricultural units based on technological differences. We fit a latent class stochastic frontier model to estimate the parameters of an output distance function formulation of the production technology to investigate if a technological separation is meaningful along sub-disciplinary lines. We apply bootstrapping techniques for model validation. Our analysis relies on evaluating a unique dataset that matches information on higher educational institutions provided by the Federal Statistical Office of Germany with the bibliometric information extracted from the ISI Web of Science Database. The estimates indicate that neglecting to account for the possible existence of latent classes leads to a biased perception of efficiency. A classification into a research-focused and teaching-focused decision-making unit improves model fit compared to the pooled stochastic frontier model. Additionally, research-focused units have a higher median technical efficiency than teaching-focused units. As the research focus is more prevalent in the biology subsample an analysis not considering the potential existence of latent classes might misleadingly give the appearance of a higher mean efficiency of biology. In fact, we find no evidence of a difference in the mean technical efficiencies for German agricultural sciences and biology using the latent class model.
本文考察了德国高等教育的技术效率，同时考虑了生产技术中可能存在的异质性。我们调查了潜在类别模型是否可以根据技术差异，在生物学和农业单位样本中识别生命科学的不同子学科。我们拟合了潜在类别随机前沿模型，以估计生产技术的输出距离函数公式的参数，以调查沿着子学科线是否存在有意义的技术分离。我们应用了引导技术来验证模型。我们的分析依赖于评估一个独特的数据集，该数据集将德国联邦统计局提供的高等教育机构信息与从 ISI Web of Science 数据库提取的文献计量信息相匹配。估计结果表明，忽略潜在类别可能存在会导致对效率的偏见感知。与综合随机前沿模型相比，将研究型和教学型决策单元分类可以提高模型拟合度。此外，研究型单位的技术效率中位数高于教学型单位。由于生物学子样本中研究重点更为普遍，因此不考虑潜在类别可能存在的分析可能会误导性地给出生物学平均效率更高的假象。实际上，我们使用潜在类别模型发现，德国农业科学和生物学的平均技术效率没有差异。
