The COVID-19 pandemic has affected firms in many economies. Exploiting treatment heterogeneity, we use a difference-in-differences design to causally identify the short-run impact of COVID-19 on innovation spending in 2020 and expected innovation spending in subsequent years. Based on a representative sample of German firms, we find that negatively affected firms substantially reduced innovation expenditure not only in the first year of the pandemic (2020) but also in the two subsequent years, indicating ’Long–Covid’ effects on innovation. In 2020, innovation expenditure fell by 4.7 % due to the pandemic. In 2022, innovation spending was even 5.4 % lower compared to the counterfactual scenario without the pandemic. Firms with higher pre-treatment digital capabilities show higher innovation resilience during the pandemic. Moreover, COVID-19 leads to a decrease in innovation spending not only in firms that were strongly negatively affected by the pandemic, but also in those firms that experienced a positive demand shock from the pandemic, presumably to increase production capacity.

We develop a simple general approach to estimating parameters of production functions. Unlike previous studies that make restrictive assumptions about the functional form of the production function, our approach efficiently estimates all parameters of any production function with Hicks-neutral productivity. Our method has the same data requirements and imposes mostly the same assumptions as the commonly used control function approach but allows to estimate gross output production functions. We validate our approach through Monte Carlo simulations and demonstate its effectiveness on empirical data from the manufacturing industries in Chile and Colombia.

The Fourth Industrial Revolution (4IR) presents a major technology transformation towards a data-driven economy, associated with both high costs and potential benefits. We build and estimate a dynamic discrete technology choice model to explain a firm’s decisions to engage in the development of new 4IR or non-4IR technologies. The model accounts for the endogenous nature of these decisions and allows them to persistently affect the firm’s future productivity path. We estimate the benefits and costs of either technology using a panel data set of high-tech manufacturing firms in Germany between 2008-2016 in combination with patent information on the type of technology. We find that firms achieve a short-run average productivity increase of 7.2% from developing 4IR technologies, 5.1% from non-4IR technologies, and 8.8% from doing both. Long-run average expected benefits arising through a strongly persistent productivity process are substantially higher for 4IR (118m Euro) than for non-4IR technologies (70m Euro). However, 4IR development costs are more than double non-4IR development costs. Especially for inexperienced firms, a combination of substantially higher development costs and lower expected long-run benefits constitute a high entry barrier to starting 4IR technology development. A subsidy for 4IR technology development shifts activities from non-4IR to 4IR while increasing overall development activities.

Pricing carbon emissions increases firms’ incentives to develop innovations aimed at reducing their productions’ carbon emission intensities. This paper incorporates this mechanism in a dynamic discrete choice model of firms’ innovation decisions while differentiating between emission-reducing and non-emission-reducing innovations. I apply the model to the European Union’s Emission Trading System and estimate its parameters using administrative carbon emission data and patent information for a large set of German manufacturing firms between 2008-2017. I find that emission-reducing innovations decrease a firm’s carbon emission intensity on average by about 13.7% while simultaneously decreasing its productivity by 1.5%. In contrast, non-emission-reducing innovations increase productivity by 2.2%. Furthermore, startup costs of emission-reducing innovations are lower than those of non-emission-reducing innovations. However, the costs of maintaining emission-reducing innovation activities are substantially higher than maintaining the development of non-emission-reducing innovation. Simulating counterfactual emission price changes substantially impacts emission-reducing innovation activity while non-emission-reducing innovations stays stable.