Interfaces with Other Disciplines
Analytical debiasing of corporate cash flow forecasts

https://doi.org/10.1016/j.ejor.2014.12.035Get rights and content

Highlights

  • We analyze the quality of cash flow forecasts of an international corporation.

  • We find rectifiable biases for all business divisions (covering 34 subsidiaries).

  • Debiasing using selected statistical correction models improves forecast accuracy.

  • Learned parameters provide decision support for corporate financial controllers.

Abstract

We propose and empirically test statistical approaches to debiasing judgmental corporate cash flow forecasts. Accuracy of cash flow forecasts plays a pivotal role in corporate planning as liquidity and foreign exchange risk management are based on such forecasts. Surprisingly, to our knowledge there is no previous empirical work on the identification, statistical correction, and interpretation of prediction biases in large enterprise financial forecast data in general, and cash flow forecasting in particular. Employing a unique set of empirical forecasts delivered by 34 legal entities of a multinational corporation over a multi-year period, we compare different forecast correction techniques such as Theil’s method and approaches employing robust regression, both with various discount factors. Our findings indicate that rectifiable mean as well as regression biases exist for all business divisions of the company and that statistical correction increases forecast accuracy significantly. We show that the parameters estimated by the models for different business divisions can also be related to the characteristics of the business environment and provide valuable insights for corporate financial controllers to better understand, quantify, and feedback the biases to the forecasters aiming to systematically improve predictive accuracy over time.

Introduction

Cash flow forecasts play a pivotal role in corporate financial management tasks. For instance, the forecasts are used in liquidity management to ensure solvency and in foreign-exchange risk management to identify and hedge exposures resulting from foreign business activities. Inaccurate forecasts are an unreliable basis for corporation-wide financial plans and can lead to liquidity shortages, uncovered currency risks or increased hedging costs. In large multinational companies, cash flow forecasts are prepared for individual subsidiaries from different countries and distinct business divisions. The cash flows are predicted by local financial managers who report the forecasts to the corporation’s central finance department, where corporation-wide financial plans are derived from all delivered forecasts.

Since the cash flow forecasts are typically generated by human experts with different backgrounds, attitudes and individual forecasting procedures, the forecasts are likely to be biased by behavioral or political dimensions, overall leading to reduced forecast accuracy. Biased judgmental forecasting and decision-making are phenomena observed in many contexts. In numerous studies and laboratory experiments significant biases are found regularly in various forecasts, decreasing the accuracy and negatively influencing business performance (Leitner and Leopold-Wildburger, 2011). For instance, in a study by Lawrence, O’Connor, and Edmundson (2000), errors in sales forecasts of three manufacturing-based companies were attributed mainly to inefficiencies and biases. In another study, Enns (2002) analyzed the influence of biased and uncertain demand forecasts on production scheduling and found that biases significantly influence lateness of the delivery to the customer.

There are several empirical studies providing further strong evidence for the existence of cognitive biases in expert forecasting, such as the well-known anchoring and adjustment. Hogarth and Makridakis (1981) and Lawrence, Goodwin, O’Connor, and Oenkal (2006) provide extensive overviews of judgmental forecasting and its heuristics and biases. Although the effect can be mitigated and reduced with adequate decision support systems (George, Duffy, Ahuja, 2000, Remus, Kottemann, 1995), it is regularly observed that information provided by decision support systems is undervalued and biases can therefore only be partly removed (Bhandari, Hassanein, Deaves, 2008, Lim, O’Connor, 1996). Overall, it is likely that corporate cash flow forecasts in multinational companies are also biased and that these biases influence decision models.

For instance, Gormley and Meade (2007) consider the problem of corporate-wide cash balance management using short-term cash flow forecasts. Based on data from a large international company, the authors show that transactional costs strongly depend on the accuracy of the cash flow forecasts used as input for financial planning.

Unfortunately, while in general there is awareness of the importance of accurate financial forecasts for corporate planning and control (Graham, Harvey, 2001, Kim, Mauer, Sherman, 1998), there is practically no research available that empirically analyzes corporate cash flow forecasts. This is particularly true regarding the identification, interpretation, and finally mitigation of judgmental biases in forecasts. Hence, corporate financial controllers have little guidance on how to assess and improve the quality of their heterogeneous cash flow forecasts. In our context, identification and removal of biases before the forecasts are used in corporation-wide planning activities could lead to increased financial efficacy.

In this paper we conduct an empirical analysis of statistical approaches to debiasing cash flow forecasts using a unique set of corporate cash flow forecasts with different horizons. The data are provided by a large international corporation and generated by experts from 34 legal entities in various countries over a period of six years and for several currencies. We analyze and discuss the impact of statistical correction methods for different estimation techniques and parameters.

To our knowledge, this is the first work to decompose, quantify, and correct cash flow forecast biases in corporate settings. More generally, we are not aware of any empirical work analyzing biases in large sets of heterogeneous enterprise financial forecast data. More important, we are not aware of scientific publications that analyze the parameters learned by correction models, although these might provide valuable insights.

Our research makes several contributions to the literature. First, we find that different types of biases exist in the empirical cash flow forecast data of our sample company. Second, we find that substantial improvements of forecast accuracy can be achieved by debiasing forecasts using statistical techniques. Third, we analyze and compare the parameters learned by the correction models and find significant differences between the respective parameters learned in different business environments. We show that the learned parameters can be related to the characteristics of the business the forecast has been generated in and argue that the parameters provide valuable insights for corporate financial controllers to better understand, quantify, feedback, and systematically mitigate biases over time.

Section snippets

Forecast debiasing methods

Many studies report successful debiasing of expert forecasts with statistical correction techniques. The most common approach is Theil’s method (Theil, 1966), which is based on a decomposition of the mean squared error (MSE). For a time series of length T with actuals A1, …, AT and corresponding forecasts F1, …, FT, the metric is defined as MSE=1T(AtFt)2.

With means F¯,A¯ and standard deviations sF, sA of the forecasts and actuals, and r as the correlation between forecasts and actuals, the

Empirical data and research design

In this section, we characterize our sample company and the available dataset before we describe the design of our empirical analysis. Results are presented subsequently in Section 4.

Experimental results

Aggregated results are presented in Table 4. Each row in the table corresponds to one of the correction techniques and parameterizations (henceforth we will refer to a technique-parameter combination as correction model or simply model). The table shows median ΔAPE achieved with a model over all forecasts and individually for each of the six division/invoice type subsamples ({AP, HP, IM} crossed with {Invoice Received (IR), Invoice Issued (II)}). The rows are ranked descending by median ΔAPE in

Summary and conclusion

In this paper, we empirically analyzed statistical error correction methods for cash flows forecasts generated by individual experts in the context of corporate financial controlling. While there is awareness of the importance of accurate financial forecasts for corporate planning and control, there is practically no research available that empirically analyzes corporate cash flow forecasts. Furthermore, there is no previous work on the selection and parameterization of error correction

References (35)

  • GormleyF.M. et al.

    The utility of cash flow forecasts in the management of corporate cash balances

    European Journal of Operational Research

    (2007)
  • GrahamJ.R. et al.

    The theory and practice of corporate finance: Evidence from the field

    Journal of Financial Economics

    (2001)
  • HarveyN.

    Why are judgements less consistent in less predictable task situations?

    Organizational Behavior and Human Decision Processes

    (1995)
  • LawrenceM. et al.

    Scale, randomness and the calibration of judgmental confidence intervals

    Organizational Behavior and Human Decision Processes

    (1993)
  • LawrenceM. et al.

    A field study of sales forecasting accuracy and processes

    European Journal of Operational Research

    (2000)
  • LeitnerJ. et al.

    Experiments on forecasting behavior with several sources of information—A review of the literature

    European Journal of Operational Research

    (2011)
  • LimJ.S. et al.

    Judgmental forecasting with interactive forecasting support systems

    Decision Support Systems

    (1996)
  • Cited by (12)

    • Fast evaluation of crack growth path using time series forecasting

      2019, Engineering Fracture Mechanics
      Citation Excerpt :

      In this study, we promote a new approach for forecasting the crack propagation based on machine learning algorithms. The main idea is to rely on time series forecasting which is a crucial issue in many disciplines [65,25,28,57,31,26,56,41,59,5,4,8,36], because the future values can be explored from its past ones with the lowest devised error. Forecasting problems based on machine learning have been successfully applied in various areas such as the economy, finance and hydrology.

    • News-based forecasts of macroeconomic indicators: A semantic path model for interpretable predictions

      2019, European Journal of Operational Research
      Citation Excerpt :

      Hence, it is likely that these forecasts Numerous studies have analyzed professional forecasts for their predictive performance and ability to identify potential biases (Blanc & Setzer, 2015; Mostard, Teunter, & de Koster, 2011). In the field of macroeconomic forecasting, a recent study (Dovern & Weisser, 2011) finds that the distribution of a forecasts accuracy varies significantly across indicators, forecasters and nations.

    • How do turbulent sectoral conditions sector influence the value of coal mining enterprises? Perspectives from the Central-Eastern Europe coal mining industry

      2018, Resources Policy
      Citation Excerpt :

      In light of this fact and in order to contribute to the general theory of value management, this article was written from the perspective of the coal mining industry. A majority of value measurement methods involve the financial profits of an enterprise (Bąk and Sierpińska-Sawicz, 2016; Kumar, 2016; Blasco and Ribal, 2013; Blanc and Setzer, 2015; Jennergren, 2008). The accounting approach uses data from balance sheets on net assets, but the value of the assets and their changes in time directly depend on the profit amount (Kowalska-Styczeń and Owczarek, 2016; Qu and Zhang, 2015; Kraus and Strömsten, 2012; Brodny, 2012).

    • Feeding-Back Error Patterns to Stimulate Self-Reflection versus Automated Debiasing of Judgments

      2023, Proceedings of the Annual Hawaii International Conference on System Sciences
    View all citing articles on Scopus
    View full text