Michael Wolf
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Michael Wolf has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
Olivier Ledoit
14 shared publicationsJoseph P. Romano
Stanford University
10 shared publicationsAzeem Shaikh
Stanford University - Department of Economics
3 shared publicationsDan Wunderli
Swiss National Bank Bern Switzerland
3 shared publications- Olivier Ledoit
University of Zürich
3 shared publications Ashok Kaul
Saarland University, Department of Economics, 66123 Saarbrucken, Germany
2 shared publications
Latest Publications
- Open Access
- 21 September 2026
- Published by Elsevier in SSRN Electronic Journal
Inference-Based Performance Evaluation Using the Sortino Ratio
The Sharpe ratio remains the most widely used measure of risk-adjusted performance, but in many practical settings the Sortino ratio is more appropriate because it focuses exclusively on downside risk-the type of risk investors care about most. For industry practitioners and applied researchers who rely on the Sortino ratio, we develop statistical methods for three fundamental inference problems: inference for the difference between two Sortino ratios; inference for a single Sortino ratio, and multiple testing based on Sortino ratios. Although sample estimates often suggest meaningful differences in performance, valid conclusions about population quantities require appropriate statistical inference. Our proposed methods accommodate key empirical features of financial returns, including non-normality, skewness, heavy tails, and serial dependence. We illustrate their practical relevance through several empirical applications based on real financial data.
- 21 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessInference-Based Performance Evaluation Using the Sortino Ratio
The Sharpe ratio remains the most widely used measure of risk-adjusted performance, but in many practical settings the Sortino ratio is more appropriate because it focuses exclusively on downside risk-the type of risk investors care about most. For industry practitioners and applied researchers who rely on the Sortino ratio, we develop statistical methods for three fundamental inference problems: inference for the difference between two Sortino ratios; inference for a single Sortino ratio, and multiple testing based on Sortino ratios. Although sample estimates often suggest meaningful differences in performance, valid conclusions about population quantities require appropriate statistical inference. Our proposed methods accommodate key empirical features of financial returns, including non-normality, skewness, heavy tails, and serial dependence. We illustrate their practical relevance through several empirical applications based on real financial data.
- Open Access
- 24 March 2026
- Published by Springer Nature in Empirical Economics
- Vol. 70 (4) , 63
The permutation test for event studies with a small number of events
Return event studies typically involve a large number of event instances. In some applications, however, this number may be very small–sometimes as few as two event instances. In such cases, standard approaches to testing average abnormal returns (AAR) or cumulative average abnormal returns (CAAR) are less effective, or may not apply at all, as they rely on central limit theorems that require large sample sizes. We propose a nonparametric permutation test that remains valid for arbitrarily small numbers of event instances. Its performance is evaluated via Monte Carlo studies, and the method is further illustrated using two empirical applications.
- 24 March 2026
- Published by Springer Nature in Empirical Economics
- Vol. 70 (4) , 63
Open AccessThe permutation test for event studies with a small number of events
Return event studies typically involve a large number of event instances. In some applications, however, this number may be very small–sometimes as few as two event instances. In such cases, standard approaches to testing average abnormal returns (AAR) or cumulative average abnormal returns (CAAR) are less effective, or may not apply at all, as they rely on central limit theorems that require large sample sizes. We propose a nonparametric permutation test that remains valid for arbitrarily small numbers of event instances. Its performance is evaluated via Monte Carlo studies, and the method is further illustrated using two empirical applications.
- Top Cited PapersOpen Access
- 22 January 2026
- Published by Springer Nature in Empirical Economics
- Vol. 70 (2) , 23
Forecasting inflation with the hedged random forest
Accurate inflation forecasting is essential for economic policy, financial markets, and broader societal stability. In recent years, machine learning methods, particularly the random forest, have shown strong potential for improving forecasting accuracy, often outperforming traditional benchmarks. Building on this success, our paper adapts the hedged random forest (HRF) of Beck et al. (The hedged random forest, 2024) to the task of forecasting inflation. Unlike the standard random forest, the HRF assigns non-equal, and potentially negative, weights to individual trees to enhance forecasting performance. We develop customized estimators for the HRF’s two key inputs: the mean and covariance matrix of the vector of forecast errors corresponding to individual trees. An extensive empirical analysis demonstrates that our approach consistently outperforms the standard random forest.
- 22 January 2026
- Published by Springer Nature in Empirical Economics
- Vol. 70 (2) , 23
Top Cited PapersOpen AccessForecasting inflation with the hedged random forest
Accurate inflation forecasting is essential for economic policy, financial markets, and broader societal stability. In recent years, machine learning methods, particularly the random forest, have shown strong potential for improving forecasting accuracy, often outperforming traditional benchmarks. Building on this success, our paper adapts the hedged random forest (HRF) of Beck et al. (The hedged random forest, 2024) to the task of forecasting inflation. Unlike the standard random forest, the HRF assigns non-equal, and potentially negative, weights to individual trees to enhance forecasting performance. We develop customized estimators for the HRF’s two key inputs: the mean and covariance matrix of the vector of forecast errors corresponding to individual trees. An extensive empirical analysis demonstrates that our approach consistently outperforms the standard random forest.
- Open Access
- 13 January 2025
- Published by Elsevier in The Quarterly Review of Economics and Finance
- Vol. 100
Markowitz portfolios under transaction costs
- 13 January 2025
- Published by Elsevier in The Quarterly Review of Economics and Finance
- Vol. 100
Open AccessMarkowitz portfolios under transaction costs
- 21 September 2024
- Published by Oxford University Press (OUP) in Journal of Financial Econometrics
- Vol. 23 (2)
Comments on: “Bootstrap Inference for Group Factor Models”
Andreou et al. (2019), henceforth AGGR (2019), propose an asymptotic test for the number of common factors, , among large panels with group-specific factors. Their test considers the one-sided testing problem ... AGGR (2019) propose a test statistic whose asymptotic distribution under the null is standard normal, which allows for the computation of a normal-theory P-value. Crucially, the test statistic involves an estimated bias correction and an estimated variance, both of which terms are not trivial to derive. Gonçalves et al. (2025), henceforth GKP (2025), propose an alternative test based on the bootstrap where the motivation is twofold. First, the test of AGGR (2019) does not allow for time-series dependence or cross-sectional dependence (conditional on the factors) of the error terms of the group-specific factor models; hence, one motivation is to provide a test that is more generally valid. Second, even when the assumptions of the normal-theory test are satisfied, it can be quite liberal in finite samples (meaning that the null-rejection probability (NRP) can be quite a bit larger than the nominal significance level); hence, another motivation of the article is to provide a test with better finite-sample properties, where the focus is on controlling the NRP.- 21 September 2024
- Published by Oxford University Press (OUP) in Journal of Financial Econometrics
- Vol. 23 (2)
Comments on: “Bootstrap Inference for Group Factor Models”
Andreou et al. (2019), henceforth AGGR (2019), propose an asymptotic test for the number of common factors, , among large panels with group-specific factors. Their test considers the one-sided testing problem ... AGGR (2019) propose a test statistic whose asymptotic distribution under the null is standard normal, which allows for the computation of a normal-theory P-value. Crucially, the test statistic involves an estimated bias correction and an estimated variance, both of which terms are not trivial to derive. Gonçalves et al. (2025), henceforth GKP (2025), propose an alternative test based on the bootstrap where the motivation is twofold. First, the test of AGGR (2019) does not allow for time-series dependence or cross-sectional dependence (conditional on the factors) of the error terms of the group-specific factor models; hence, one motivation is to provide a test that is more generally valid. Second, even when the assumptions of the normal-theory test are satisfied, it can be quite liberal in finite samples (meaning that the null-rejection probability (NRP) can be quite a bit larger than the nominal significance level); hence, another motivation of the article is to provide a test with better finite-sample properties, where the focus is on controlling the NRP.- Open Access
- 23 November 2023
- Published by Springer Nature in Empirical Economics
- Vol. 66 (6) , 2435-2450
Single-firm inference in event studies via the permutation test
Return event studies generally involve several firms but there are also cases when only one firm is involved. This makes the relevant testing problems, abnormal return and cumulative abnormal return, more difficult since one cannot exploit the multitude of firms (by using a relevant central limit theorem, say) to design hypothesis tests. We propose a permutation test which is of nonparametric nature and more generally valid than the tests that have previously been proposed in the literature in this context. We address the question of the power of the test via a brief simulation study and also illustrate the method with two applications to real data.
- 23 November 2023
- Published by Springer Nature in Empirical Economics
- Vol. 66 (6) , 2435-2450
Open AccessSingle-firm inference in event studies via the permutation test
Return event studies generally involve several firms but there are also cases when only one firm is involved. This makes the relevant testing problems, abnormal return and cumulative abnormal return, more difficult since one cannot exploit the multitude of firms (by using a relevant central limit theorem, say) to design hypothesis tests. We propose a permutation test which is of nonparametric nature and more generally valid than the tests that have previously been proposed in the literature in this context. We address the question of the power of the test via a brief simulation study and also illustrate the method with two applications to real data.
- Open Access
Hedging Forecast Combinations With an Application to the Random Forest
This papers proposes a generic, high-level methodology for generating forecast combinations that would deliver the optimal linearly combined forecast in terms of the mean-squared forecast error if one had access to two population quantities: the mean vector and the covariance matrix of the vector of individual forecast errors. We point out that this problem is identical to a mean-variance portfolio construction problem, in which portfolio weights correspond to forecast combination weights. We allow negative forecast weights and interpret such weights as hedging over and under estimation risks across estimators. This interpretation follows directly as an implication of the portfolio analogy. We demonstrate our method's improved out-of-sample performance relative to standard methods in combining tree forecasts to form weighted random forests in 14 data sets.
Open AccessHedging Forecast Combinations With an Application to the Random Forest
This papers proposes a generic, high-level methodology for generating forecast combinations that would deliver the optimal linearly combined forecast in terms of the mean-squared forecast error if one had access to two population quantities: the mean vector and the covariance matrix of the vector of individual forecast errors. We point out that this problem is identical to a mean-variance portfolio construction problem, in which portfolio weights correspond to forecast combination weights. We allow negative forecast weights and interpret such weights as hedging over and under estimation risks across estimators. This interpretation follows directly as an implication of the portfolio analogy. We demonstrate our method's improved out-of-sample performance relative to standard methods in combining tree forecasts to form weighted random forests in 14 data sets.
- Open Access
- 1 July 2019
- Published by Elsevier in SSRN Electronic Journal
Shrinkage Estimation of Large Covariance Matrices: Keep it Simple, Statistician?
Under rotation-equivariant decision theory, sample covariance matrix eigenvalues can be optimally shrunk by recombining sample eigenvectors with a (potentially nonlinear) function of the unobservable population covariance matrix. The optimal shape of this function reflects the loss/risk that is to be minimized. We introduce a broad family of covariance matrix estimators that can handle all regular functional transformations of the population covariance matrix under large-dimensional asymptotics. We solve the problem of optimal covariance matrix estimation under a variety of loss functions motivated by statistical precedent, probability theory, and differential geometry. The key statistical ingredient of our nonlinear shrinkage methodology is a new estimator of the angle between sample and population eigenvectors, without making strong assumptions on the population eigenvalues. We also compare our methodology to two simpler ones from the literature, linear shrinkage and shrinkage based on the spiked covariance model, via both Monte Carlo simulations and an empirical application.
- 1 July 2019
- Published by Elsevier in SSRN Electronic Journal
Open AccessShrinkage Estimation of Large Covariance Matrices: Keep it Simple, Statistician?
Under rotation-equivariant decision theory, sample covariance matrix eigenvalues can be optimally shrunk by recombining sample eigenvectors with a (potentially nonlinear) function of the unobservable population covariance matrix. The optimal shape of this function reflects the loss/risk that is to be minimized. We introduce a broad family of covariance matrix estimators that can handle all regular functional transformations of the population covariance matrix under large-dimensional asymptotics. We solve the problem of optimal covariance matrix estimation under a variety of loss functions motivated by statistical precedent, probability theory, and differential geometry. The key statistical ingredient of our nonlinear shrinkage methodology is a new estimator of the angle between sample and population eigenvectors, without making strong assumptions on the population eigenvalues. We also compare our methodology to two simpler ones from the literature, linear shrinkage and shrinkage based on the spiked covariance model, via both Monte Carlo simulations and an empirical application.
- Open Access
- 1 May 2019
- Published by Elsevier in SSRN Electronic Journal
The Power of (Non-)Linear Shrinking: A Review and Guide to Covariance Matrix Estimation
Many econometric and data-science applications require a reliable estimate of the covariance matrix, such as Markowitz portfolio selection. When the number of variables is of the same magnitude as the number of observations, this constitutes a difficult estimation problem; the sample covariance matrix certainly will not do. In this paper, we review our work in this area going back 15+ years. We have promoted various shrinkage estimators, which can be classified into linear and nonlinear. Linear shrinkage is simpler to understand, to derive, and to implement. But nonlinear shrinkage can deliver another level of performance improvement, especially if overlaid with stylized facts such as time-varying co-volatility or factor models.
- 1 May 2019
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Power of (Non-)Linear Shrinking: A Review and Guide to Covariance Matrix Estimation
Many econometric and data-science applications require a reliable estimate of the covariance matrix, such as Markowitz portfolio selection. When the number of variables is of the same magnitude as the number of observations, this constitutes a difficult estimation problem; the sample covariance matrix certainly will not do. In this paper, we review our work in this area going back 15+ years. We have promoted various shrinkage estimators, which can be classified into linear and nonlinear. Linear shrinkage is simpler to understand, to derive, and to implement. But nonlinear shrinkage can deliver another level of performance improvement, especially if overlaid with stylized facts such as time-varying co-volatility or factor models.
- Open Access
- 6 February 2019
- Published by Wiley in Stem Cells International
- Vol. 2019, 1-7
Scalable Culture Strategies for the Expansion of Patient-Derived Cancer Stem Cell Lines
Cancer stem cells (CSCs) have recently raised great interest as a promising biological system for designing effective cancer therapies. The scarcity of CSCs in vivo and the consequent low numbers obtained from biopsies represent a major hurdle to the development of such strategies. It is therefore necessary to design robust scalable methods to enable efficient expansion of bona fide CSCs in vitro. Here, we evaluated the applicability of computer-controlled bioreactors combined with 3D aggregate culture and microcarrier technology, widely used in stem cell bioprocessing, for the expansion and enrichment of CSCs isolated from different types of solid tumors—colorectal cancer (CRC) and non-small-cell lung cancer (NSCLC) from two patients. Results show that these culture strategies improved cell expansion and CSC enrichment. Both patient-derived CSC lines were able to grow on microcarriers, the best results being achieved for PPlus 102-L, Pro-F 102-L, Fact 102-L, and CGEN 102-L beads (5-fold and 40-fold increase in total cell concentration for CRC and NSCLC cells, respectively, in 6 days). As for 3D aggregate culture strategy, the cell proliferation profile was donor dependent. NSCLC cells were the only cells able to form aggregates and proliferate, and the flat-bottom bioreactor vessel equipped with a trapezoid-shaped paddle impeller was the most efficient configuration for cell growth (21-fold increase in cell concentration achieved in 8 days). Serum-free medium promotes CSC enrichment in both 3D aggregate and microcarrier cultures. The protocols developed herein for CSC expansion have the potential to be transferred to clinical and industrial settings, providing key insights to guide bioprocess design towards the production of enriched CSC cultures in higher quantity and improved quality.
- 6 February 2019
- Published by Wiley in Stem Cells International
- Vol. 2019, 1-7
Open AccessScalable Culture Strategies for the Expansion of Patient-Derived Cancer Stem Cell Lines
Cancer stem cells (CSCs) have recently raised great interest as a promising biological system for designing effective cancer therapies. The scarcity of CSCs in vivo and the consequent low numbers obtained from biopsies represent a major hurdle to the development of such strategies. It is therefore necessary to design robust scalable methods to enable efficient expansion of bona fide CSCs in vitro. Here, we evaluated the applicability of computer-controlled bioreactors combined with 3D aggregate culture and microcarrier technology, widely used in stem cell bioprocessing, for the expansion and enrichment of CSCs isolated from different types of solid tumors—colorectal cancer (CRC) and non-small-cell lung cancer (NSCLC) from two patients. Results show that these culture strategies improved cell expansion and CSC enrichment. Both patient-derived CSC lines were able to grow on microcarriers, the best results being achieved for PPlus 102-L, Pro-F 102-L, Fact 102-L, and CGEN 102-L beads (5-fold and 40-fold increase in total cell concentration for CRC and NSCLC cells, respectively, in 6 days). As for 3D aggregate culture strategy, the cell proliferation profile was donor dependent. NSCLC cells were the only cells able to form aggregates and proliferate, and the flat-bottom bioreactor vessel equipped with a trapezoid-shaped paddle impeller was the most efficient configuration for cell growth (21-fold increase in cell concentration achieved in 8 days). Serum-free medium promotes CSC enrichment in both 3D aggregate and microcarrier cultures. The protocols developed herein for CSC expansion have the potential to be transferred to clinical and industrial settings, providing key insights to guide bioprocess design towards the production of enriched CSC cultures in higher quantity and improved quality.