Testing for multiple change-points in macroeconometrics: an empirical guide and recent developments

Otilia Boldea, Tilburg University, Alastair R. Hall, University of Manchester · 2025

Read Testing for multiple change-points in macroeconometrics: an empirical guide and recent developments by Otilia Boldea, Tilburg University, Alastair R. Hall, University of Manchester (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-22204.txt.

Introduction

Otilia Boldea and Alastair R Hall wrote this guide for a handbook on hands-on macroeconomics, due in a second edition. Their topic is change-points: moments when the pattern in economic data shifts, such as a new slope in a trend. Finding several such breaks at once is hard, and older guides often treat only shifts in average level.

The pair review fresh methods for models with outside drivers, for panel data that tracks many units, and for factor models. They stress tests that find the breaks one by one, in order. The chapter sticks to methods built on the broad assumptions that real economic work needs, so users can apply them with care.

Catalog record

CREATOROtilia Boldea, Tilburg University, Alastair R. Hall, University of Manchester
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-22204

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Testing for multiple change-points in macroeconometrics: an empirical guide and recent developments Thanks: This manuscript is prepared for the second edition of Handbook of Research Methods and Applications on Empirical Macroeconomics , edited by N. Hashimzade and M. Thornton, to be published by Edward Elgar. Otilia Boldea, Tilburg University, Alastair R. Hall, University of Manchester Abstract We review recent developments in detecting and estimating multiple change-points in time series models with exogenous and endogenous regressors, panel data models, and factor models. This review differs from others in multiple ways: (1) it focuses on inference about the change-points in slope parameters , rather than in the mean of the dependent variable - the latter being common in the statistical literature; (2) it focuses on detecting - via sequential testing and other methods - multiple change-points, and only discusses one change-point when methods for multiple change-points are not available; (3) it is meant as a practitioner’s guide for empirical macroeconomists first, and as a result, it focuses only on the methods derived under the most general assumptions relevant to macroeconomic applications. JEL classification codes: C12, C22, C23, C26, C32, C33, C36 Keywords: change-points, break-points, time series, panel data, factor models, sequential testing, information criteria, lasso 1 Introduction Parametric models for macroeconomic time series involve the assumption that the specified relationship remains the same over the entire sample, a property referred to as structural stability . However, since the earliest days of macroeconometric analysis, researchers have been concerned about the appropriateness of the assumption that such a property holds over long periods of time; for example see ? . This concern is also central to the so-called Lucas (1976) critique which has played a central role in shaping macroeconometric analysis. ? emphasizes the fact that the decision models of economic agents are hard to describe in terms of stable parameterizations, simply because changes in policy may change these decision models and their respective parametrization. These arguments underscore the importance of using tests of structural stability as diagnostic checks for macroeconometric models. A large body of empirical macroeconomic studies provides evidence for parameter change in a variety of macroeconomic models. For example, there is compelling evidence that the New Keynesian Phillips curve flattened or changed over time - ? , ? , ? , ? , ? , ? . Similarly, there is evidence that the relationship between policy interest rates and inflation/output changed over time, shifting from output targeting to inflation targeting, or to dual output and inflation targeting, and to a zero lower bound regime more recently - ? , ? , ? , ? , ? . There is also strong evidence that the response of many key macroeconomic variables to fiscal and monetary policies changes over the sample and over the state of the economy - see e.g. ? , ? , ? , ? , ? . Further examples of parameters changing over time can be found for growth, output, exchange rate and unemployment models. 1 1 1 See e.g. ? for examples. If these parameter changes are ignored in the estimation procedure, they lead to incorrect policy recommendations and flawed macroeconomic forecasts. Thus, it is essential - and it has become common practice - to test parameter instability in macroeconomic models. Parameter changes can be of many types. In this chapter, we describe tests for discrete change that is, scenarios in which the parameters change abruptly at particular points in time. These points tend to be known as break-points in the econometrics literature and as change-points in the statistics literature; and we use these terms interchangeably. In some cases, researchers are interested in testing whether change is associated with specific events. For example, ? investigate whether monetary policy policy reaction function of the Federal Reserve Board is different during the tenure of different chairmen. In such circumstances the break-points are specified as part of the test, and said to be known (or fixed). While there is an obvious appeal to addressing this type of question, the outcome of such tests must be interpreted with caution because a significant statistic may reflect parameter change at a different date. For this reason, it has become common to treat the break-points as unknown that is, the dates of potential change are not specified as part of the test. Furthermore, there are typically many events that potentially cause parameter change and so the ideal statistical framework encompasses the case in which neither the number nor the location (in time) of the break-points is known. To date, this ideal has been developed in the context of several linear model types employed in macroeconometrics, such as univariate and multivariate regressions, panel data regressions and linear factor models. In these settings, the number and location of these change-points can be estimated via sequential testing procedures. 2 2 2 Note that these models are linear only in-between change-points, as change-points induce an abrupt change that can be perceived as a nonlinearity. For nonlinear models, results are more limited, being confined in the most general scenarios - with endogenous regressors - to the case of a single break-point. In linear models, a sequential testing strategy is computationally attractive because it can be performed using the dynamic programming algorithm developed by ? . However, this inferential approach does not deliver the correct number of change-points as the sample size increases unless the significance level of each test in the procedure shrinks to zero as the sample size increases. This problem can be mitigated by controlling the size of the whole sequential procedure via a Bonferroni correction. Other inferential approaches are available that, under the right conditions, may retrieve the correct number of change-points, but may be computationally more intensive. Examples of such methods are information criteria and penalized regressions, such as adaptive lasso and mixed-integer programming. This chapter contains discussion of some of these other non-test-based procedures, but does not explore them in the same depth as the sequential testing approach due to space constraints. The literature on testing and estimating multiple change-points is vast, and spans across many areas of science and social science, from econometrics and economics to finance, statistics, environmental science, information systems, machine learning, engineering, health, to mention just a few. In the last decade, several new reviews and books have been written summarizing various change-point detection methods - see reviews by ? , ? , ? , ? , ? , and the book by ? . This chapter differs from all these in multiple ways: (1) it focuses on inference about the change-points in slope parameters , rather than in the mean of the dependent variable - the latter being common in the statistical literature; (2) it focuses on detecting - via sequential testing and other methods - multiple change-points, and only discusses one change-point when methods for multiple change-points are not available; (3) it is meant as a practitioner’s guide for empirical macroeconomists first, and as a result, it focuses only on the methods derived under the most general assumptions relevant to macroeconomic applications, including autocorrelated errors when possible. However, because it is a practical guide first, we do not formalize the results with asymptotics; interested readers are referred to the

Read the complete text