Mark Kritzman
Research Subjects
Mark Kritzman has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- David Turkington
Is senior managing director and head of State Street Associates in Cambridge, MA
20 shared publications - Megan Czasonis
Is a managing director at State Street Associates in Cambridge, MA
12 shared publications - Will Kinlaw
Is a senior managing director at State Street Global Markets in Boston, MA
4 shared publications - Baykan Pamir
Is an assistant vice president at State Street Associates in Cambridge, MA
2 shared publications - Baykan Pamir
Is a junior trader at Alphachain Capital in London, UK
2 shared publications A. Ang
1 shared publication
Latest Publications
- 29 August 2026
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 8 (4) , 9-24
Confidence Revisited: The Distribution of Information
Classical statistics assesses the confidence of a prediction by measuring error variance, estimation variance, or the sum of both. More complex machine learning models assess confidence using empirical techniques, such as resampling or conformal prediction. These methods implicitly assume that the model is correctly specified and the predictive variables are measured without error. The authors describe a model-free approach to prediction, called relevance-based prediction, which does not rely on a model’s correctness or assume that the predictive variables are error-free. It measures the uncertainty of each individual prediction based on the unique distribution of information that is used to form the prediction.
- 29 August 2026
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 8 (4) , 9-24
Confidence Revisited: The Distribution of Information
Classical statistics assesses the confidence of a prediction by measuring error variance, estimation variance, or the sum of both. More complex machine learning models assess confidence using empirical techniques, such as resampling or conformal prediction. These methods implicitly assume that the model is correctly specified and the predictive variables are measured without error. The authors describe a model-free approach to prediction, called relevance-based prediction, which does not rely on a model’s correctness or assume that the predictive variables are error-free. It measures the uncertainty of each individual prediction based on the unique distribution of information that is used to form the prediction.
- 20 March 2026
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 52 (7) , 114-123
Policy Regimes versus Economic Regimes: Implications for Portfolio Construction
Starting during the Global Financial Crisis and for several years thereafter, then again during COVID, zero-interest rate policy (ZIRP) reshaped asset returns and portfolio construction. As rates normalized, asset managers faced a practical challenge: while ZIRP and non-ZIRP periods are distinct policy regimes, they do not cleanly map to distinct economic environments defined by inflation, growth, and unemployment. This ambiguity complicates strategic asset allocation because institutional investors who had treated regimes as binary occurrences must now approach regimes in a more nuanced way. We introduce a regime-aware approach that measures the “regimeness” of historical periods using a statistic called relevance, allowing allocators to weight past observations by their alignment with prevailing economic conditions rather than by rigid classifications. Using this framework, we generate regime-dependent forecasts of expected returns, volatilities, and correlations across major asset classes and construct portfolios for post-ZIRP conditions. We find that explicitly incorporating regime ambiguity yields clearer allocation guidance, including materially different positioning across credit and commodities. Beyond asset allocation, the framework provides a transparent and adaptive structure for institutional decision-making in environments where policy signals do not neatly define economic reality.
- 20 March 2026
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 52 (7) , 114-123
Policy Regimes versus Economic Regimes: Implications for Portfolio Construction
Starting during the Global Financial Crisis and for several years thereafter, then again during COVID, zero-interest rate policy (ZIRP) reshaped asset returns and portfolio construction. As rates normalized, asset managers faced a practical challenge: while ZIRP and non-ZIRP periods are distinct policy regimes, they do not cleanly map to distinct economic environments defined by inflation, growth, and unemployment. This ambiguity complicates strategic asset allocation because institutional investors who had treated regimes as binary occurrences must now approach regimes in a more nuanced way. We introduce a regime-aware approach that measures the “regimeness” of historical periods using a statistic called relevance, allowing allocators to weight past observations by their alignment with prevailing economic conditions rather than by rigid classifications. Using this framework, we generate regime-dependent forecasts of expected returns, volatilities, and correlations across major asset classes and construct portfolios for post-ZIRP conditions. We find that explicitly incorporating regime ambiguity yields clearer allocation guidance, including materially different positioning across credit and commodities. Beyond asset allocation, the framework provides a transparent and adaptive structure for institutional decision-making in environments where policy signals do not neatly define economic reality.
- 10 September 2025
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 7 (4) , 8-21
Prediction with Incomplete Information
A key requirement for forming data-driven predictions is to assemble the best possible set of observations for the predictive variables. This task, however, is not often easy. In the case of time-series data, some variables have shorter histories than others, and some variables are reported less frequently than others. And in the case of cross-sectional data, some information is not reported for every case. We are therefore faced with several choices. We can discard predictive variables with missing information. We can exclude observations with missing information and retain only those with full information for all the predictive variables. We can use statistical techniques to manufacture replacements for the missing information. Each of these approaches, however, has significant drawbacks. The authors propose a new procedure for treating missing information that enables us to retain more information than model-based approaches and, at the same time, to account for the relative reliability of observations with missing information.
- 10 September 2025
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 7 (4) , 8-21
Prediction with Incomplete Information
A key requirement for forming data-driven predictions is to assemble the best possible set of observations for the predictive variables. This task, however, is not often easy. In the case of time-series data, some variables have shorter histories than others, and some variables are reported less frequently than others. And in the case of cross-sectional data, some information is not reported for every case. We are therefore faced with several choices. We can discard predictive variables with missing information. We can exclude observations with missing information and retain only those with full information for all the predictive variables. We can use statistical techniques to manufacture replacements for the missing information. Each of these approaches, however, has significant drawbacks. The authors propose a new procedure for treating missing information that enables us to retain more information than model-based approaches and, at the same time, to account for the relative reliability of observations with missing information.
- 18 February 2024
- Published by With Intelligence LLC in The Journal of Portfolio Management
Dynamic Warp Analysis: A New Approach for Detecting and Timing Bubbles
The authors apply a technique called dynamic warp analysis to rescale the unique cadences of 2,638 bubbles into synchronized steps. They then observe the distributions of chosen stock characteristics for each step across all the bubbles. They also observe these stock characteristics during periods when a stock is not experiencing a bubble. The authors use this information to detect when a bubble is under way and how far it has progressed. They test several trading rules to assess the potential to profit from this information.
- 18 February 2024
- Published by With Intelligence LLC in The Journal of Portfolio Management
Dynamic Warp Analysis: A New Approach for Detecting and Timing Bubbles
The authors apply a technique called dynamic warp analysis to rescale the unique cadences of 2,638 bubbles into synchronized steps. They then observe the distributions of chosen stock characteristics for each step across all the bubbles. They also observe these stock characteristics during periods when a stock is not experiencing a bubble. The authors use this information to detect when a bubble is under way and how far it has progressed. They test several trading rules to assess the potential to profit from this information.
- 6 September 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 50 (1) , 8-18
Portfolio Construction When Regimes Are Ambiguous
Investors sometimes have strong convictions that a distinctive economic regime will prevail in the period ahead and therefore would like to form a portfolio that reflects the expected returns, standard deviations, and correlations of assets during such a regime. To do so, they typically isolate a subsample of returns in which a regime indicator, such as the rate of economic growth, is above or below a chosen threshold and estimate expected returns, standard deviations, and correlations by equally weighting the observations within the subsample. This approach assumes that every observation within the regime subsample is equally important to forming the estimates, regardless of whether an observation coincides with a growth rate that is far from the threshold or one that is only marginally distant from the threshold. Moreover, with this approach it is problematic to describe a regime by more than a single indicator because there is no nonarbitrary way to combine the indicators and because the addition of indicators increases the likelihood of producing an empty or overly sparse subsample. The authors apply a new concept called relevance to estimate regime-specific expected returns, standard deviations, and correlations. Their relevance-based approach explicitly accounts for the importance of an observation to forming an estimate, and it seamlessly enables the inclusion of multiple regime indicators in a principled way.
- 6 September 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 50 (1) , 8-18
Portfolio Construction When Regimes Are Ambiguous
Investors sometimes have strong convictions that a distinctive economic regime will prevail in the period ahead and therefore would like to form a portfolio that reflects the expected returns, standard deviations, and correlations of assets during such a regime. To do so, they typically isolate a subsample of returns in which a regime indicator, such as the rate of economic growth, is above or below a chosen threshold and estimate expected returns, standard deviations, and correlations by equally weighting the observations within the subsample. This approach assumes that every observation within the regime subsample is equally important to forming the estimates, regardless of whether an observation coincides with a growth rate that is far from the threshold or one that is only marginally distant from the threshold. Moreover, with this approach it is problematic to describe a regime by more than a single indicator because there is no nonarbitrary way to combine the indicators and because the addition of indicators increases the likelihood of producing an empty or overly sparse subsample. The authors apply a new concept called relevance to estimate regime-specific expected returns, standard deviations, and correlations. Their relevance-based approach explicitly accounts for the importance of an observation to forming an estimate, and it seamlessly enables the inclusion of multiple regime indicators in a principled way.
- 28 June 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 49 (9) , 96-104
An Intuitive Guide to Relevance-Based Prediction
Relevance-based prediction is a new approach to data-driven forecasting that serves as a favorable alternative to both linear regression analysis and machine learning. It follows from two seminal scientific innovations: Prasanta Mahalanobis’ distance measure and Claude Shannon’s information theory. Relevance-based prediction rests on three key tenets: 1) relevance, which measures the importance of an observation to a prediction; 2) fit, which measures the reliability of each individual prediction task; and 3) codependence, which holds that the choice of observations and predictive variables should be determined jointly for each individual prediction task.
- 28 June 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 49 (9) , 96-104
An Intuitive Guide to Relevance-Based Prediction
Relevance-based prediction is a new approach to data-driven forecasting that serves as a favorable alternative to both linear regression analysis and machine learning. It follows from two seminal scientific innovations: Prasanta Mahalanobis’ distance measure and Claude Shannon’s information theory. Relevance-based prediction rests on three key tenets: 1) relevance, which measures the importance of an observation to a prediction; 2) fit, which measures the reliability of each individual prediction task; and 3) codependence, which holds that the choice of observations and predictive variables should be determined jointly for each individual prediction task.
- 10 May 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 49 (7) , 81-92
Event Time
Investors take for granted that returns are recorded in units of time, such as days, months, and years. Yet some time periods include unusual events that reasonably cause asset prices to change, whereas other periods are relatively free of unusual events, in which case returns mostly reflect noise. Based on insights from information theory, the authors rescale time into event units so that each return is related to a common degree of event intensity. Their analysis reveals that when returns are measured in event units, their distributions are more normal and their co-occurrences are more stable, which enables analysts to form more reliable inferences.
- 10 May 2023
- Published by With Intelligence LLC in The Journal of Portfolio Management
- Vol. 49 (7) , 81-92
Event Time
Investors take for granted that returns are recorded in units of time, such as days, months, and years. Yet some time periods include unusual events that reasonably cause asset prices to change, whereas other periods are relatively free of unusual events, in which case returns mostly reflect noise. Based on insights from information theory, the authors rescale time into event units so that each return is related to a common degree of event intensity. Their analysis reveals that when returns are measured in event units, their distributions are more normal and their co-occurrences are more stable, which enables analysts to form more reliable inferences.
- 1 December 2022
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 5 (1) , 27-46
Relevance-Based Prediction: A Transparent and Adaptive Alternative to Machine Learning
The authors describe a new prediction system based on relevance, which gives a mathematically precise measure of the importance of an observation to forming a prediction, as well as fit, which measures a specific prediction’s reliability. They show how their relevance-based approach to prediction identifies the optimal combination of observations and predictive variables for any given prediction task, thereby presenting a unified alternative to both kernel regression and lasso regression, which they call CKT regression. They argue that their new prediction system addresses complexities that are beyond the capacity of linear regression analysis but in a way that is more transparent, more flexible, and less arbitrary than widely used machine learning algorithms.
- 1 December 2022
- Published by With Intelligence LLC in The Journal of Financial Data Science
- Vol. 5 (1) , 27-46
Relevance-Based Prediction: A Transparent and Adaptive Alternative to Machine Learning
The authors describe a new prediction system based on relevance, which gives a mathematically precise measure of the importance of an observation to forming a prediction, as well as fit, which measures a specific prediction’s reliability. They show how their relevance-based approach to prediction identifies the optimal combination of observations and predictive variables for any given prediction task, thereby presenting a unified alternative to both kernel regression and lasso regression, which they call CKT regression. They argue that their new prediction system addresses complexities that are beyond the capacity of linear regression analysis but in a way that is more transparent, more flexible, and less arbitrary than widely used machine learning algorithms.
- 3 May 2022
- Published by With Intelligence LLC in Practical Applications
Practical Applications of The Role of Cryptocurrencies in Investor Portfolios
In The Role of Cryptocurrencies in Investor Portfolios, from the Spring 2022 issue of The Journal of Alternative Investments, Megan Czasonis of State Street Associates, Mark Kritzman of the MIT Sloan School of Management and Windham Capital Management, Baykan Pamir of Alphachain Capital, and David Turkington of State Street Associates examine correlations among asset class returns, with a focus on those of stocks and cryptocurrencies. They highlight that correlation is time-varying and dependent on the frequency of data. They find that cryptocurrencies produce less of a diversification benefit than other asset classes such as bonds or gold. Their analysis provides a template for investigating the role of cryptocurrencies in investor portfolios as the market develops. The approach provides more relevant information than a single, full-sample correlation number.
- 3 May 2022
- Published by With Intelligence LLC in Practical Applications
Practical Applications of The Role of Cryptocurrencies in Investor Portfolios
In The Role of Cryptocurrencies in Investor Portfolios, from the Spring 2022 issue of The Journal of Alternative Investments, Megan Czasonis of State Street Associates, Mark Kritzman of the MIT Sloan School of Management and Windham Capital Management, Baykan Pamir of Alphachain Capital, and David Turkington of State Street Associates examine correlations among asset class returns, with a focus on those of stocks and cryptocurrencies. They highlight that correlation is time-varying and dependent on the frequency of data. They find that cryptocurrencies produce less of a diversification benefit than other asset classes such as bonds or gold. Their analysis provides a template for investigating the role of cryptocurrencies in investor portfolios as the market develops. The approach provides more relevant information than a single, full-sample correlation number.
- 23 February 2022
- Published by With Intelligence LLC in The Journal of Alternative Investments
- Vol. 24 (4) , 18-32
The Role of Cryptocurrencies in Investor Portfolios
The role of cryptocurrencies as a vehicle for speculation has been well established. However, it is less clear if cryptocurrencies can also serve to manage risk. The authors seek to determine the diversification potential of cryptocurrencies both for short and long horizons. For short horizons, they estimate correlations that consider the direction and magnitude of returns for relevant asset classes, rather than focusing on full-sample correlations, as is customary. For long horizons, they compute “single period correlations” that capture the extent to which cryptocurrencies move synchronously with, or drift apart from, other assets over an investor’s horizon. They also identify utility-maximizing allocations to cryptocurrencies directly from historical return samples that account for all features of the data as well as more nuanced preferences than are typically assumed.
- 23 February 2022
- Published by With Intelligence LLC in The Journal of Alternative Investments
- Vol. 24 (4) , 18-32
The Role of Cryptocurrencies in Investor Portfolios
The role of cryptocurrencies as a vehicle for speculation has been well established. However, it is less clear if cryptocurrencies can also serve to manage risk. The authors seek to determine the diversification potential of cryptocurrencies both for short and long horizons. For short horizons, they estimate correlations that consider the direction and magnitude of returns for relevant asset classes, rather than focusing on full-sample correlations, as is customary. For long horizons, they compute “single period correlations” that capture the extent to which cryptocurrencies move synchronously with, or drift apart from, other assets over an investor’s horizon. They also identify utility-maximizing allocations to cryptocurrencies directly from historical return samples that account for all features of the data as well as more nuanced preferences than are typically assumed.