I have been working as an Economist in the Structural and Climate Policies Division of the Research Department at the International Monetary Fund (IMF) since July 2024. My research fields are macroeconomics, labor economics, and public economics.
My current research focuses on technological change, structural reforms, emerging skill demand, and AI diffusion, with particular emphasis on their macroeconomic and labor-market implications. My work draws on macro- and micro-level data, including administrative employer-employee records, online job postings, and survey data. At the IMF, I also contribute to policy analysis and support country teams.
I am further interested in understanding how economic policies and institutions shape household decisions, labor-market outcomes, and inequality. My work in this area examines the redistributive role of childcare subsidies or the implications of divorce expectations for household decisions and gender earnings inequality.
My research has been featured in various media outlets, including The Economist, the Financial Times, Bloomberg, the World Economic Forum, The New York Times, and Die Welt.
Prior to joining the IMF, I was a Postdoctoral Scholar at the University of Mannheim and a member of the Collaborative Research Center TR 224 “Economic Perspectives on Societal Challenges” in project A03 with Michèle Tertilt. I also worked as a Research Associate at the European University Institute (EUI) with Philipp Kircher on his ERC-Project “Technological Change: New Sources, Consequences, and Impact Mitigation”.
I obtained a PhD in Economics at the EUI in September 2020. My supervisors were Árpád Ábrahám and Dominik Sachs.
Here is my CV.
Contact: dk.economist at pm.me
New Working Paper, November 2026: Leading Firms and the Future of Work
New Working Paper, October 2026: Optimistic Divorce Beliefs and the Life-Cycle Economics of the Family
Revise & Resubmit at AEJ: Economic Policy with Quantifying Okun’s Leaky Bucket: The Case of Progressive Childcare Subsidies, October 2026
Updated Working Paper, August 2026: Quantifying Okun’s Leaky Bucket: The Case of Progressive Childcare Subsidies
IMF Staff Discussion Note, January 2026: Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age
Understanding how occupations will rise or fall with technological and organizational change over the next decade is essential for labor market adjustment. This paper develops an approach to forecasting occupational change by identifying leading firms whose workforce composition anticipates the future occupational structure of their industry as other firms catch up over time. We develop a theoretical model with technology diffusion and frontier-laggard dynamics showing that leading firms are early adopters of technology and how this can be exploited for forecasting. We apply its insights by (1) machine-learning current characteristics of firms whose current occupational structure is close to the future occupational structure of its industry, and (2) estimating a forecasting model that combines current and lagged occupational structures of industries and leaders. Application in a hold-out sample a decade later shows substantial improvements in predictions relative to a model without leaders. We outperform official forecasts and predictions based on automation indices. The machine-learned leader forecast for 2023 correlates negatively with several measures of automation–a relationship with reverses for early automation measures when projecting out to 2033, while staying negative for recent AI- and GPT-based measures.
Mergers and acquisitions reshape not only who employs workers but also which skills firms seek. Using U.S. job postings linked to acquisitions (2010–2024), we show that acquirers ask less often for frontier skills after a deal. Each completed deal lowers the share of an acquirer’s postings requiring emerging IT skills by about half of its pre-deal level; demand for established skills does not change. The decline is concentrated in cross-industry deals between firms whose pre-deal postings sought similar skills; there, acquirers nearly stop posting for frontier skills. These patterns suggest that acquisitions can partly substitute for open-market recruiting of frontier skills.
Nearly half of U.S. marriages end in divorce, yet spouses plan as if theirs will not. Our survey shows married individuals underestimate their divorce risk against three benchmarks. We embed these beliefs in a life-cycle model with endogenous divorce, fertility, and learning-by-doing human capital. Optimistic couples specialize more and have children earlier, leaving secondary earners with less human capital, which persists after divorce. Correcting beliefs closes about 6% of the gender earnings gap and 7% of the long-run child penalty. The expected welfare cost of optimism is small but concentrated among failed marriages.
We formalise and estimate the dynamic marginal efficiency cost of redistribution (MECR) in the spirit of Okun’s “leaky bucket” to compare income-contingent childcare subsidies with the broader tax-transfer system. Because the MECR depends on heterogeneous behavioral responses and fiscal effects, we estimate a structural model of household childcare demand and maternal labor supply for Germany, allowing for informal childcare and leisure. Our analysis identifies two competing forces: labor supply responses increase the MECR of childcare subsidies relative to the tax-transfer system, while child development effects decrease it. Under empirically plausible assumptions, progressive childcare subsidies are the more efficient redistribution tool.
In the United States, 35–40% of all marriages end in divorce. Yet, we provide survey evidence that, on average, married respondents expect a divorce likelihood of 15%, with most respondents significantly underestimating their predicted divorce risk. Our survey reveals that individuals with more overoptimistic divorce expectations exhibit higher within-couple inequality in market hours and earnings and accumulate significantly less wealth than their rational counterparts. Building on this evidence, we incorporate overoptimistic divorce expectations into a household life-cycle model with endogenous accumulation of human capital, assets, and ex-ante heterogeneity in spouses’ wages. Couples jointly choose their market hours, home production, and joint savings. We quantify the model using U.S. microdata and show that overoptimism leads to (1) higher within-couple specialization and (2) lower savings, as overoptimistic lower-wage spouses fail to internalize the insurance value of human capital and assets in the event of divorce. Overoptimism during marriage persists beyond divorce through lower assets and human capital upon divorce, with particularly adverse consequences for the lower-wage spouse. The model thus provides a novel explanation for the high poverty rates observed among divorced mothers. Finally, we show that joint taxation of married couples amplifies specialization among overoptimistic couples.
We study how labor markets price newly emerging skills within jobs. Using U.S. online postings from Lightcast between 2010 and 2024, we identify “new skills” as those rarely requested in 2010–11 but prevalent thereafter. New skill demands—including those that are AI–related—are heavily concentrated in computer and analytical domains across occupations, with little emphasis on routine work. Comparing vacancies within the same detailed occupation, industry, firm, and county in a given year and conditioning on the overall number of listed skills, we find that postings listing at least a new skill offer 2–4 percent higher wages; the premium increases steeply with the number of new skills and is concentrated in IT skills. Notably, new skills related to developing AI tools carry a premium of about 9 percent, compared with about 4 percent for new skills linked to using AI tools. In contrast to new skills, new job titles are a weaker signal of market value: new-title postings do not carry a wage premium by itself but only if paired with new skills. Linking new skills to upstream breakthrough patents yields substantially larger returns—a finding reinforced by instrumental variable estimates—tying wage premia to frontier technological innovation.
This paper provides causal evidence on the impact of subsidy re-allocation between high technology sectors and low-skill sectors on local labor markets. We exploit a policy targeting under-performing employment areas, France’s Aides à Finalité Régionale, which relaxes rules governing the allocation of firm subsidies while keeping their level constant. In response, policy makers re-allocate subsidies away from research and development to mainly low-skilled manufacturing and service sectors. It triggers a persistent improvement of employment, mainly through increased low-skilled manufacturing employment and at the expense of R&D related occupations. In the long term, though, labor income and productivity decrease. Finally, at the individual level, workers employed in manufacturing at the time of the treatment benefit on average of 2% higher hourly wage even 10 years after the policy was lifted.
Linking data from the National Longitudinal Survey of Youth 1979 (NLSY79) and the NLSY79 Children and Young Adults, we document a substantial positive correlation of employment status between mothers and their offspring in the United States. After controlling for ability, education, fertility and wealth, offspring of permanently employed mothers have an 11 percentage-point higher probability to be employed in each given year than those of never employed mothers. The intergenerational transmission of maternal employment is stronger to daughters but significant also to sons. Investigating potential mechanisms, we provide suggestive evidence for a role model channel, through which labor force participation may be transmitted. Offspring seem to emulate the example of their mother when they observe her working. By contrast, we are able to rule out alternative candidate explanations such as network effects, occupation-specific human capital and local conditions of the labor market.
We examine the impact of the global recession triggered by the Covid-19 pandemic on women’s versus men’s employment. Whereas recent recessions in advanced economies had a disproportionate impact on men’s employment, giving rise to the moniker “mancessions,” we show that the pandemic recession of 2020 was a “shecession” with larger employment declines among women in most countries. We examine the causes behind this pattern using micro data from several national labor force surveys, and show that both the composition of women’s employment across industries and occupations as well as increased childcare needs during closures of schools and daycare centers made important contributions. Gender gaps in the employment impact of the pandemic arise almost entirely among workers who are unable to work from home. Among telecommuters a different kind of gender gap arises: women working from home during the pandemic spent more work time also doing childcare and experienced greater productivity reductions than men. We identify two key challenges for future research. First, why is the pandemic gender gap pervasive, i.e., why did women experience larger employment reductions than men even after accounting for industry/occupation and childcare effects? Second, how will the pandemic shape gender equality in a post-pandemic labor market that will likely continue to be characterized by pervasive telecommuting?
This note synthesizes insights from a high-level workshop and scenario-planning exercise on the global economic and financial implications of artificial intelligence (AI) hosted by the IMF in December 2025. AI is advancing rapidly and has the potential to restructure the global economy, which calls for treating AI as a macro-critical transition rather than a standard technology shock. However, the macroeconomic path will be shaped less by frontier capability alone than by the speed and breadth of diffusion and the readiness of institutions and infrastructure to absorb the technology. Understanding AI’s potential and implications is crucial for the Fund’s ability to support its member countries with high-quality policy advice.
Since 2016, Saudi Arabia’s Vision 2030 reforms have improved governance, business regulations, capital markets, the labor market, and the external sector, narrowing structural gaps with frontier economies and improving economic performance. This paper summarizes progress, estimates the output impact of the reforms, notes challenges, and highlights future priorities for continued growth and diversification.
We formalize and estimate the dynamic marginal efficiency cost of redistribution (MECR) in the spirit of Okun’s “leaky bucket” to compare the MECR of an income-contingent childcare subsidy program and of the income-contingent tax and transfer schedule. We set up a dynamic structural model of heterogeneous households choosing their childcare demand and maternal labor supply. Allowing for the availability of informal childcare and for consumption of leisure, we estimate this model within the German context. Our analysis identifies two competing forces. (i) Labor supply responses increase the MECR of the childcare subsidy relative to the tax and transfer system. (ii) Child development effects decrease the MECR of the childcare subsidy relative to the income tax. We show that, under most plausible assumptions on the long-term returns to childcare attendance for children growing up in households of different incomes, progressive childcare subsidies are the more efficient redistribution tool.