On the rising uncertainties and deviations from rules

When monetary policies are reviewed historically, it would not be wrong to say that the basic conditions of effective monetary policies are transparency and predictability. Although these two concepts seem to be simple at first glance, there are very few examples where these two conditions are fully provided. Transparency is a concept that can be explained mostly by the functioning of democratic institutions, whereas predictability is quite complicated. A group of economists, including Milton Friedman and John Taylor, suggest that monetary policy rules are an important tool for providing predictability. Although there is strong evidence that this hypothesis is correct when the economic data are examined, there are alternative ideas claiming that simple mathematical equations may be unable to explain the functioning of the chaotic world. Additionally, the strongest argument presented against monetary policy rules is the acceptance of the natural unemployment hypothesis. However, in the empirical literature, there is evidence that there is no fixed natural unemployment rate, and this rate is constantly changing. Accordingly, the relationship between inflation and unemployment is variable (Lear, 2000).

When the monetary policies implemented by the FED in the historical process are reviewed from a broad perspective, especially the sharp change at the beginning of the twenty-first century was quite likely to bring new perspectives on the “Rules vs Discretion” discussion. In the early 2000s, it was observed that the monetary policies of the FED did not correspond to the popular monetary policy rules. Considering the political and economic power of the USA, it would not be expected that these deviations in the FED’s policies would only have a local effect. As expected, the monetary authorities in other countries, especially the ECB, followed these deviations in the FED’s policies (Gray, 2013). Policy deviations are a phenomenon experienced throughout Europe, but the size of the deviations differed between countries since inflation and economic growth rates were not equal in all European countries (Thakor, 2015). In the meantime, a fluctuation in the US housing market deepened further and became a global economic crisis. According to Taylor (2009), the easy-money monetary policy implemented by the FED prior to the crisis contributed to the deepening of the crisis. Considering that the effects of the local economic crises that occurred in the late 20th century were limited, the relationship between the deviations in the monetary policy rules and the economic crises brought about a big question mark. The main arguments of the economists who asserted that there was a correlation between these two phenomena were not different from the main arguments of “Rules vs Discretion” discussions, suggesting that monetary policy rules reduced uncertainty in the market and paved the way for economic actors to make more rational decisions. Counter-arguments claimed that, due to the economic conjuncture, monetary authorities deviated from monetary policies to prevent deterioration in macroeconomic and financial indicators.

Undoubtedly, this study does not make assertive promises to end a debate that has been continuing for many years. However, it contains significant findings in order to provide a different perspective. This article aims to investigate the causality relationship between monetary policy deviations and economic and financial uncertainties in general terms. In this respect, the monetary policies of the FED are classified under two headings as rule-based period and discretionary period, as is often done in the literature. The results indicate that the causality relationship between monetary policy deviations and uncertainty indicators is weak in cases where monetary policies can be explained with certain rules.

“Rules vs Discretion” is one of the most popular discussions in the history of monetary policy. It may seem like a simple question at first glance, but there is a highly important intellectual background and deep literature behind it. Although the historical roots of monetary policy rules reach back to Smith (1776) and Ricardo (1824), the formation of modern monetary policy rules dates back to the first half of the twentieth century. The second half of the twentieth century was the period when discussions on monetary policy started to ripen. It would not be wrong to say that especially the studies by Milton Friedman (1958, 1960) and Henry Simons form the basis of today’s monetary policy theories. According to these two economists, monetary authorities aim at continuous growth in money supply via open market operations to stabilize the aggregate demand. Their main purpose was to protect the economy against unexpected shocks resulting from discretionary monetary policies and prevent negative situations such as inflation, recession, and crisis. From this perspective, these two economists can be considered reformists in monetary policy discussions.

Studies by Kydland and Prescott (1977) and Barro and Gordon (1983) are two of the most important studies of the period after Friedman and Simons. The argument in these two studies is that a very fundamental problem, such as the time inconsistency in discretionary monetary policies, arises. Investors and entrepreneurs specify their economic position at time t according to the policy makers’ commitment at time t−1. However, considering the changing global and national economic dynamics, these commitments are not fulfilled most of the time. The problem of time inconsistency constitutes quite a significant part of the literature (Davis, Fujiwara, & Wang, 2018; Surico, 2008; Ireland, 1999; Calvo, 1978).

The economic literature shaped around the “Rules vs Discretion” discussion is not as sharply divided as the topic of discussion itself. Although popular monetary policy rules suggested by economists, such as Taylor (1993) and McCallum (1988), are theoretically presented as a reference to monetary authorities, it is not possible in practice to define monetary policies precisely with mathematical equations. The view of the economists who currently advocate rule-based policies is that discretionary policies will cause instability in the markets and therefore monetary policies should be explained with certain mathematical equations. In this context, many economists criticized the interventionist policies of the FED, especially in the 2008 crisis, and provided important evidence that the crisis was deepened due to these interventions (Cecchetti, Flores-Lagunes, & Krause, 2006; Kahn, 2010; Taylor, 2012).

One of the significant studies on this topic, Kahn (2010), argues that deviations from the Taylor rule influence financial conditions in some sectors, although this effect is not equally pronounced across all markets. Additionally, it provides compelling evidence that keeping short-term interest rates low for an extended period encouraged a housing price bubble and led investors to seek riskier, higher-yielding investments. In a similar study, Ahrend (2010) concludes that periods during which short-term interest rates are kept significantly below those predicted by the Taylor rule led to increases in housing prices, creating imbalances in the market. Another prominent study, Taylor (2013), argues that deviations from monetary policy rules have spillover effects across countries, disrupt the balance of the international financial system, and consequently negatively impact economic performance. On the other hand, Piazzesi (2014) emphasizes that standard loss functions in inflation and output are lower under rule-based monetary policies, but it is unclear whether the Taylor rule had a positive impact on economic performance during the 2008 financial crisis.

When the recent literature on this topic is reviewed, it is observed that, despite the use of large datasets and different econometric methods, the results remain similar. Teryoshin (2023) obtained strong empirical evidence that rule-based monetary policies are more effective in maintaining price stability compared to discretionary monetary policies, using a very large dataset. In a different perspective, Ohanyan and Grigoryan (2021) argue that there is a significant correlation between discretionary policies and volatility in financial markets.

In this study, the Taylor rule, which is often used in the economic literature, was taken as a reference, and monetary policy deviations were calculated over this equation. In short, the Taylor rule aims to determine interest rates by focusing on inflation and output, as in Equations (1) or (2) (Taylor, 1993);

$i$ represents the policy interest rate, $p$ represents the inflation rate compared to the previous four quarters, and $y$ represents the output gap ^[1]. As per this rule, central banks should aim to minimize the output gap by keeping inflation at the targeted rate. The central banks should determine the interest rates according to deviations that may occur in the variables. This rule is simple and reasonable and has the power to explain the behaviour of the FED within a certain date range. After John Taylor’s publication of the study, the Taylor rule soon attracted considerable attention both nationally and internationally (Mankiw, 2009, p. 416). There is highly important economic literature on the effect of deviations of policy interest rates from the Taylor rule on both macroeconomic variables (Nikolsko-Rzhevskyy, Papell, & Prodan, 2014; Ince, Molodtsova, & Papell, 2016; Wilde, 2012; Anderl & Caporale, 2022) and housing prices (Fitwi, Hein, & Mercer, 2015; O'Meara, 2015; Morley & Wei, 2012). Many questions such as “Can the behaviours of central banks be explained by the Taylor rule?” and “Is the Taylor rule a usable monetary policy?” were examined by several researchers to form the basis of their studies (Castro, 2011; Papadamou, Sidiropoulos, & Vidra, 2018; Beju & Ciupac-Ulici, 2015).

Similar to Taylor’s (2012) study, this article examines the monetary policy of the USA from 1985 to the present in two periods. The part from 1985 to 2003 is called the rule-based period, and the part from 2004 to the present is called the discretionary period. The FED’s orientation to rule-based policies started during Paul Volcker’s period (1979–1987). As seen in Figure 1, it can be said that the monetary policies of the FED can be explained by the Taylor rule, especially in Volcker’s period. The crises of the early 90s (oil price shock) and 2000s (dot-com bubble and the September 11 terrorist attacks) are quite important opportunities to examine the rule-based period. When the deviations from the Taylor rule during these periods are carefully reviewed, it is seen that the deviations are quite small, and the monetary policies implemented by the FED fit into a certain framework. The impression that monetary authorities” decisions on the monetary policy are based on a certain plan gives confidence to investors and entrepreneurs, and this situation quickly eliminates the monetary policy-related uncertainties. It can be said that one of the most important reasons for bypassing the crises experienced in the rule-based period with less damage is the trust of investors and capital owners in the monetary authorities.

The period after 2003, during which deviations from policy rules increased, is very different from the previous period, and this period can be examined in general within the scope of the causes and consequences of the 2008 global crisis. The most evident difference in the Figure 2 is that the deviations from the policy rules are extremely high. Undoubtedly, it would be a superficial assumption to say that these deviations are the single cause of economic crises and recessions. However, the fact that sharp changes in the monetary policy were not openly shared with the public substantially damaged the principles of transparency and predictability, which economic policies are based on. Moreover, the unusual interventions of the FED in the markets during the 2008 crisis were quite critical. Practices such as Term Auction Facility (TAF) and the Troubled Asset Relief Program (TARP), which were particularly implemented to save financial institutions and the companies they owed to, could not display the expected effects in the markets; on the contrary, operational inconsistencies caused uncertainty in the markets (Taylor, 2012).

The smoothing method is the most used method in the de-trending of time series. It is quite common to calculate the output gap, especially by de-trending the output variable. There are many different methods in the process of calculating the output gap. In this study, frequently used methods such as Hodrick-Prescott (HP), Christiano-Fitzgerald (CF), and Exponential Regression were selected.

The Exponential Regression equation, the first method used, is considered the simplest method to calculate the potential output and can be defined as in Equations (3), (4), and (5).

When the natural logarithm of both sides of the equation is taken;

When the error term is added;

The Hodrick-Prescott (HP) filter, the second method, is the most popular method for calculating the output gap in the literature. As seen in Equations (6) and (7), HP estimates the potential output by minimizing the trend and cyclicality of the present output variable.

$gt$ represents the trend component, $λ$ represents the smoothing parameter that penalizes the volatility in the trend. The main purpose of the HP filter is to estimate a flattened $g$ trend from its $c$ series. There is a correlation between $λ$ and the trend. As $λ$ approaches infinity, the trend becomes linear. However, when the literature on HP is reviewed, it can be seen that there is no definite consensus on the value of $λ$⁠. In this study, the value of $λ$ was chosen as 1,600 based on the study by Ravn and Uhlig (2002).

The last of the smoothing methods is the Christiano-Fitzgerald (CF) filter, which is a generalized version of the Baxter-King filter. The filter proposed by Christiano and Fitzgerald (2003) is based on the solution of the minimization problem in Equation (9) for a finite set.

As shown in Figure 3, the exponential regression method reflected shocks more clearly compared to the other two methods. Furthermore, there is a high correlation between the output gaps calculated by the HP and CF methods.

To be able to examine the causality relationship between the variables, the nonlinear form of the VAR models proposed by Sims (1980) as an alternative to simultaneous equation systems was used. The classical linear VAR model can be expressed as in Equation (10).

$Xt$ represents the kx1 dimensional vector of the internal variables, $εt$ (⁠$εt∼iid)$ refers to the error term, $β$ denotes the parameters estimated with the least squares, and $p$ represents the lag length. Time series can exhibit nonlinear behaviours in some cases. In these cases, using nonlinear models provides more accurate results. The two-regime nonlinear VAR model based on the Self-Exciting Threshold Autoregressive (SETAR) model can be defined as in Equations (11) and (12).

The variables represented by $Xt$ and $εt$ are the same as those in the linear VAR model. Moreover, $It$ represents the indicator function, $φt−d$ denotes the threshold value variable, and $p$ refers to the threshold value. The parameters can be estimated by minimizing the function number (13) with the least squares method.

The data used in the study were received from the Federal Reserve Bank of St. Louis, and as mentioned in the previous sections, the data set was divided into two as rule-based period and discretionary period. As seen in Table 1, containing the descriptive statistics of the data, deviations and uncertainties are high on the dates defined as the discretionary period, whereas federal funds rates are low. Six different variables were used in the estimated models. The variables used can be studied under two subheadings as monetary policy deviations and economic and financial indicators. Variables in the first group represent the output gaps calculated by three different methods, as demonstrated in the previous sections. Among the variables in the second group, Economic Policy Uncertainty (EPU) represents monetary policy uncertainty, while Equity Market Volatility (EMV) and National Financial Conditions Index (NFCI) represent changes in financial markets. Three different models were established to examine the interaction between the variables.

The EPU variable, which was included in the model to represent economic policy uncertainties, is an index mainly based on the news in newspapers. The explanatory power of this index, which was first created by Baker, Bloom, and Davis (2016) for 12 countries and is often used in empirical studies, is quite high. It is mainly formed by taking the frequency of the news that contains words representing categories such as “uncertainty,” “economics,” and “politics” in leading US newspapers as a reference. The methodology used by Baker et al. (2016) was adapted to the economy of many countries after the study became popular (Cerda, Silva, & Valente, 2016; Zalla, 2017; Ghirelli, Perez, & Urtasun, 2019). As in the EPU variable, the EMV variable was created by Baker, Bloom, Davis, and Kost (2019) on the basis of the news in newspapers. Similar to the EPU, this index was created by taking the frequency of the selected keywords as a reference. The selected keywords can be examined under three headings: E′ (economic, economy, financial), M′ (stock market, equity market, Standard and Poors), and V′ (Volatility, Realized Volatility, Uncertainty, Risk, VIX). The NFCI data created by the Chicago Fed are shared to provide information about the financial conditions in monetary markets, debt and stock markets in the USA. A positive series indicates tighter financial conditions, and a negative series indicates looser financial conditions (Brave & Kelley, 2017).

The ADF and PP tests were conducted to check whether the variables used in the analysis were stationary. The findings in Table A1 show that the variables were not stationary with their level values but became stationary when their first differences were taken.

Upon reviewing the results obtained from the empirical analysis, it is seen that the power of policy deviations and economic policy uncertainty to trigger each other in the periods when rule-based monetary policies are implemented is significantly lower compared to the discretionary period (Tables 2, 3, 4). These results can be said to significantly confirm the hypotheses proposed by John Taylor and other economists, who argue that rule-based monetary policies will be more efficient. The results obtained are in line with studies such as Kahn (2010), Ahrend (2010), and Piazzesi (2014), which examine the effects of deviations from monetary policy rules on various economic variables. While other studies in the literature that obtain similar results generally focus on the direct effects of deviations from monetary policy rules on financial markets and housing prices, this study examines the effects of these deviations on selected economic and financial indices. This approach aims to provide a broader analysis of economic and financial impacts beyond the scope of existing research.

One of the most important study’s findings is that the causality relationship between policy deviations and economic policy uncertainty disappeared to a large extent due to increasing uncertainties in periods when monetary policies could not be explained by mathematical equations. In other words, economic actors assume that high uncertainties are permanent in periods when a certain threshold is crossed and shape their decisions according to this assumption.

In light of all these findings, it can be said that the likelihood of unexpected shocks that occur in the periods when rule-based policies are implemented to trigger crises is lower compared to the period when discretionary policies are implemented. The measures taken by the FED to prevent the 2008 crisis can be evaluated in this respect. Measures such as TARP and TAF, which were taken to prevent the crisis during periods when the FED diverged from policy rules and kept interest rates low, did not yield results as expected. Undoubtedly, one of the reasons for this situation is the undecided image exhibited by the FED during these interventions. Considering the causality relationship between the variables, it is likely that such interventions will yield more positive results in periods when rule-based monetary policies are implemented, although it is against their nature.

In addition to all these, adopting a specific monetary policy rule and making decisions in line with that rule under all circumstances seem particularly challenging in today’s world, where globalization is on the rise and the economy is highly dependent on numerous internal dynamics. Therefore, it is crucial for monetary authorities to enhance their communication strategies. Explaining deviations to the public along with their justifications would significantly help prevent these deviations from triggering crises and uncertainties. In particular, in developing economies where monetary authorities have not yet gained international credibility, such communication strategies are highly significant.

The robustness of the findings obtained in the study was tested with different methods. First, the output gap variable in the equations referring to the Taylor rule (Equations (1) and (2)) was calculated with three different methods, and the results were compared. When the models are compared, no significant difference is seen between the output gap calculation methods. Although the results are similar, the calculation methods that best explain the monetary policies implemented by the FED in both the rule-based period and the discretionary period are the HP and CF methods. Unlike these methods, the exponential regression method overreacts to fluctuations, particularly in the rule-based period. Additionally, the low output gap coefficient in Equations (1) and (2) prevented the direct reflection of the differences between the calculation methods on the findings.

Moreover, the correlation matrix was examined to identify the multicollinearity problem of the predicted models. As seen in Table 1, the correlations between the variables are low (less than 80%), and no multicollinearity problem exists in the models. Furthermore, the BDS test, which is the most used test for linearity in the econometric literature, was used to examine the suitability of the series for nonlinear analysis methods. The test proposed by Broock, Scheinkman, Dechert, and LeBaron (1996) is mathematically based on the correlation integral. According to the results in Table A2, the structure of the models is suitable for nonlinear econometric methods.

Economic crises lead to serious human dramas, but they constitute a very productive research area for economists. It can easily be said that world economies have an extremely rich history in this respect. The global economy has witnessed many crises over the last 50 years because of both the internal variables of the economy and non-economic reasons. None of the fluctuations until the 2008 crisis had turned into a large-scale global crisis. For certain, it is impossible to explain the evolution of the fluctuations in the US real estate market into a global crisis similar to the Great Depression with a single variable. However, many economists present strong evidence that the monetary policies of the FED triggered this crisis. From a broader perspective, the literature on the correlation between monetary policies and economic crises has a highly large and complex structure. This study focuses more specifically on monetary policy rules instead of making a general inference about this broad framework.

The causality relationship between deviations from monetary policy rules and uncertainty in the markets has been discussed for a long time. The environment of economic and political uncertainty that has arisen in almost every region of the world in recent years has raised these discussions again. Especially after 2003, the fact that the monetary policies of the FED diverged from being defined with mathematical equations and that it tried to keep interest rates at low levels provides a unique opportunity to compare these two monetary policy strategies. The main objectives of this study can be gathered under two headings. The first is to provide a different perspective for the “rules vs discretion” discussion with empirical findings. The second is to obtain findings on the causality relationship between deviations from monetary policy rules and uncertainties. In other words, the study aims to answer the question, “Do deviations cause uncertainty or do uncertainties make deviations necessary?'.

As in all other economic questions, it is difficult to answer this question with a definite causality relationship. The findings indicate that the causality relationship between the variables evolved into different forms under different conditions. However, in general, two main findings obtained from the empirical analysis can be focused on. The first is that rule-based monetary policies prevent variables from triggering each other. One of the most important reasons for this situation is that economic actors take rational decisions in compliance with economic paradigms within the framework of predictable monetary policies. The second is that the causality relationship between the variables almost disappears in periods when deviations from the rules and uncertainties are high. This finding can be interpreted as economic actors starting to think that long-term uncertainties will be permanent if they cross a certain threshold. In other words, it can be said that the uncertainty has started to be normalized. This normalization process is quite important to understand the effects of uncertainty in economic and financial markets on the decision-making mechanisms of economic actors.

This study focuses on a popular monetary policy rule and tests the hypothesis under these constraints. In future research, during this period when central banks are moving away from conventional policies, the use of rules that incorporate unconventional monetary policy instruments may better explain current economic conditions. Moreover, new methods for calculating the output gap could increase the reliability of the findings.