Teodor Duevski

I examine how the breadth of venture capital (VC) partners' human capital influences investment selection, startup performance, and innovation. Partners with broader human capital are more likely to lead investments in novel startups with previously unexplored business models and significantly increase their likelihood of major success; however, they underperform when leading non-novel deals. Exploiting plausibly exogenous variation in partner time constraints as a shock to the within-VC firm likelihood of leading a deal, I provide causal evidence for these effects. A theoretical model endogenizes startup creation, partner assignment, and investment to rationalize the empirical findings and provide additional testable predictions. The results highlight the nuanced value of human capital breadth in financing innovation.

We show how Limited Partners' (LPs) environmental and social (E&S) concerns transmit to private equity (PE) firms through capital supply. E&S incidents in portfolio companies reduce PE fundraising, as E&S-concerned relationship LPs refrain from recommitting and are not easily substituted. Using a legal reform that expands E&S-concerned public pension capital to PE, we causally show that PE firms internalise LPs' E&S concerns, reducing dirty sector portfolio allocation and increasing ESG hiring. Additionally, PE firms with E&S-concerned relationship LPs engage with portfolio companies to manage E&S risks. Limited capital substitutability enables LPs to delegate E&S preferences, shaping PE allocation and engagement.

We revisit the measurement of capital allocation concentration in the venture capital (VC) industry and highlight the shortcomings of the widely used Herfindahl-Hirschman Index (HHI). We show that HHI-based concentration measures are sensitive to discrepancies in VC database coverage, industry taxonomies, and classification granularity, yielding divergent trends even when applied to identical VC portfolios. To overcome these limitations, we develop a novel concentration metric using large language model (LLM) text embeddings that capture the semantic similarity among financed startups beyond predefined industry classifications. Using matched PitchBook and Crunchbase data, we validate our approach and show that OpenAI embeddings outperform alternative models on signal-to-noise and retrieval tasks. Applying this methodology, we document that aggregate VC capital allocation concentration has increased more sharply than suggested by HHI measures. A novel decomposition shows that 40% of the growth in capital allocation concentration stems from an increase in within-sector similarity among founded startups, an effect that industry-based measures do not capture.

This paper examines how venture capital (VC) firms allocate capital between known markets (exploitation) and new markets (exploration). I develop a dynamic agency model with moral hazard and learning, where a VC raises funds from a limited partner (LP) and chooses its portfolio strategy. In equilibrium, firms with high opportunity costs avoid exploration, while those that explore base their decisions on fund management costs and learning ability. LPs' threat of divestment helps mitigate moral hazard and can encourage exploration. Using PitchBook data, I construct a novel measure of the opportunity cost of exploration based on VC partners' prior investment experience. Empirically, firms with higher exploration costs specialize more, perform better in their first fund, but exhibit slower growth in subsequent fundraising.

I study how product market competition shapes firms' intangible investment decisions. I extend the seminal d'Aspremont and Jacquemin (1988) framework to incorporate two forms of intangible investments: production-based (fully non-rival, excludable) and demand-shifter (partially non-rival, partially excludable) in a multi-product market setting. The model predicts that production-based investments increase with a firm's product-market span, while investment in demand-shifting intangibles follows an inverted-U pattern due to cross-market spillovers. Using a panel of U.S. public firms, I find empirical evidence consistent with these predictions. These findings highlight the importance of non-rivalry, limited excludability, and spillovers in shaping intangible investment decisions in multi-market environments.

In statistics, samples are drawn from a population in a data-generating process (DGP). Standard errors measure the uncertainty in estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence-generating process (EGP). We claim that EGP variation across researchers adds uncertainty—nonstandard errors (NSEs). We study NSEs by letting 164 teams test the same hypotheses on the same data. NSEs turn out to be sizable, but smaller for more reproducible or higher rated research. Adding peer-review stages reduces NSEs. We further find that this type of uncertainty is underestimated by participants.

Grain boundaries play a key role in the performance of thin-film optoelectronic devices and yet their effect in halide perovskite materials is still not understood. The biggest factor limiting progress is the inability to identify grain boundaries. Noncrystallographic techniques can misidentify grain boundaries, leading to conflicting literature reports about their influence; however, the gold standard—electron backscatter diffraction (EBSD)—destroys halide perovskite thin films. Here, this problem is solved by using a solid-state EBSD detector with 6000 times higher sensitivity than the traditional phosphor screen and camera. Correlating true grain size with photoluminescence lifetime, carrier diffusion length, and mobility shows that grain boundaries are not benign but have a recombination velocity of 1670 cm s−1, comparable to that of crystalline silicon. Amorphous grain boundaries are also observed that give rise to locally brighter photoluminescence intensity and longer lifetimes. This anomalous grain boundary character offers a possible explanation for the mysteriously long lifetime and record efficiency achieved in small grain halide perovskite thin films. It also suggests a new approach for passivating grain boundaries, independent of surface passivation, to lead to even better performance in optoelectronic devices.