Comparing Predictive Accuracy, Twenty Years Later: A Personal Perspective on the Use and Abuse of Diebold–Mariano Tests

By Francis X. Diebold. 2015

519 citations recorded by the source corpus.

Subfields: Economics and Econometrics, Finance

View Paper: https://www.semanticscholar.org/paper/c7e208d3b8e3f8f14366aff6233991e0f4fb6e8a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Finance

Citations in the corpus

Empirical Asset Pricing via Machine Learning (2018, unknown): “To make pairwise comparisons of methods, we use the Diebold and Mariano (1995) test for differences in out-of-sample predictive accuracy between two models.”
History and trends in solar irradiance and PV power forecasting: A preliminary assessment and review using text mining (2018, background): “author FRANCIS DIEBOLD wrote another paper discussing the widespread misuse of his test across various scientific domains (Diebold, 2015).”
Exchange Rate Predictability (2013, unknown): “64 Alternatively, the forecast evaluation exercise could target: (i) the direction of prediction, as in Engel (1994), Cheung et al.”
Machine Learning Advances for Time Series Forecasting (2020, methodology): “See also the recent discussion in Diebold (2015).”
Risk Everywhere: Modeling and Managing Volatility (2018, unknown): “Bollerslev, Diebold, and Labys (2000). Intuitively, in the absence of any contaminating market”
The role of the information set for forecasting—with applications to risk management (2014, methodology): “4 we investigate the properties of the Diebold and Mariano (1995) test in the situation of nested sequences of information sets and asymptotically ideal forecasts.”
The VIX, the Variance Premium and Stock Market Volatility (2013, unknown): “West, 2006), recent research suggests that the Diebold and Mariano test works well even in model-based out-of-sample forecasting comparisons (see Clark and McCracken, 2011; Diebold, 2013).”
Bad Environments, Good Environments: A Non-Gaussian Asymmetric Volatility Model (2015, unknown): “While we could in principle use more complex statistics that take the model structure and estimation into account, recent research suggests that the Diebold and Mariano test works well even in model-based out-of-sample forecasting comparisons (see Clark and McCracken, 2011; Diebold, 2013).”
A Kolmogorov-Smirnov Based Test for Comparing the Predictive Accuracy of Two Sets of Forecasts (2015, methodology): “The DM test can be briefly introduced as an asymptotic z-test for the hypothesis that the loss differential is zero [6] (1).”
Quadratic Variance Swap Models (2016, unknown): “…assess the statistical differences of the model pricing errors, we run various Diebold-Mariano (DM) tests.28 For each 28 We follow standard practice and use Diebold-Mariano tests to draw conclusions about models, rather than about model forecasts; see Diebold (2012) for a discussion of this point.”
Probabilistic forecasting in day-ahead electricity markets: Simulating peak and off-peak prices (2018, background): “For further information on the DM test, see Diebold and Mariano (2002) and Diebold (2015).”
On the sources of uncertainty in exchange rate predictability (2018, unknown): “First, regressions which allow for time-changing sets of predictors and varying degrees of coefficients adaptivity yield a smaller Root Mean Squared Forecast Error (RMSFE) than the RW at the (14)The Diebold and Mariano (1995) and West (1996) test is computed as: DMW = f √ P/[sample variance of f̂t+h”
Recent advances in electricity price forecasting: A review of probabilistic forecasting (2016, methodology): “is zero [1, 91], where S i(·, ·) is the score (or loss) of model i.”
Exogenous drivers of Bitcoin and Cryptocurrency volatility – A mixed data sampling approach to forecasting (2019, background): “…and compare information criteria of models with an identical number of parameters which makes a penalization irrelevant, we conduct a pseudo-out-of-sample and forecasting analysis over various time horizons to overcome this issue with the cost of less predictive power (Diebold, 2015, Hansen, 2010).”
Measuring news sentiment (2020, methodology): “In-sample tests are advantageous because they use all of the available data, however, recent studies have discussed reasons why out-of-sample tests may have better finite-sample performance (Hansen (2010), Diebold (2015)).”