Neo-Wicksellian Monetary Aggregates

The paper introduces a comparison between two macroeconomic models used for monetary policy evaluation: the standard neo-Wicksellian (NW) model and a newly developed Banks and Bonds (BB) model. The NW model, while simpler, is criticized for its incompleteness, particularly its lack of explicit modeling of banks and financial institutions. The BB model addresses these shortcomings by incorporating banks and the role of government bonds in household and bank liquidity management. The central research question is whether the simpler NW model accurately reflects macroeconomic behavior when compared to the more comprehensive BB model. The study will achieve this by analyzing second moments, variance decompositions, and impulse response functions from both models, and by examining the role of monetary aggregates and velocity in inflation prediction. This comparison aims to determine whether the increased complexity of the BB model significantly improves the accuracy of monetary policy analysis and forecasting, or if the simpler NW model offers a sufficient approximation. The study intends to test whether the often-ignored monetary aggregates in the NW framework are actually useful for inflation forecasting, a question of significant interest in macroeconomic analysis.

This section delves into the structural differences between the NW and BB models. The BB model introduces banks, making deposits and loans costly, which generates financial frictions and gives monetary aggregates a more significant role in economic activity. The model also incorporates government bonds used by banks and households for liquidity management. This creates a spread between the consumption Euler equation rate (CCAPM rate) and the government bond rate, reflecting a liquidity premium. This spread can act as a buffer against liquidity crunches from contractionary open market operations (OMOs), having implications for interest rate surprises and the interest rate rule itself. In contrast, the NW model is embedded within the BB model. It strips out banks, assumes only cash is used for transactions, and therefore lacks the financial frictions and liquidity premium features of the BB model. This difference forms the basis of the comparison, focusing on the degree to which simplifying assumptions of the NW model lead to inaccuracies in representing actual macroeconomic behavior and informing monetary policy.

The models are calibrated using quarterly averages of inflation, federal funds rates, and government bond-to-GDP ratios. The paper then compares the unconditional moments, impulse response functions, and variance decompositions generated by the two models. The focus is on the quantitative significance of the theoretical differences. While the BB model might be a more complete representation of the economy, the study investigates whether the simpler NW model provides a reasonably accurate approximation. The comparison involves analyzing how various shocks (interest rate, productivity, government spending, and money demand) pass through each model economy. The analysis also examines whether the NW model underestimates the usefulness of monetary aggregates in inflation forecasting, employing artificial datasets and bivariate VARs between inflation and monetary aggregate growth. The role of government spending shocks, and their implications for inflation under both models, are examined in detail, drawing attention to differences that may arise from the structural features of each model. The paper explores how these differences are interpreted in terms of the ‘liquidity buffering’ and ‘liquidity provision’ effects.

This section discusses the implications of the model comparison for monetary policy and suggests avenues for future research. It explores the predictive power of various monetary aggregates in inflation forecasting. The findings reveal that while simple monetary aggregates like growth in nominal cash balances are not useful in the NW model, broader measures like M2 velocity and total government liabilities (dLIAB) show predictive power in the BB model, suggesting that more comprehensive models might better capture the complex interplay between financial markets and inflation. The research also touches upon the issue of model indeterminacy when the Taylor principle is violated, showing that the non-Ricardian nature of the BB model, arising from the role of government bonds in liquidity management, eliminates this problem, unlike the NW model. The conclusion highlights the importance of considering the liquidity buffering and liquidity provision effects in monetary policy analysis. It also points towards the necessity of more detailed research into the role of bank loans and financial frictions, which may impact the results. This section emphasizes the need for further investigation into the medium-term predictive power of the identified monetary aggregates, to provide a more robust assessment of their usefulness in monetary policy decision-making.

The BB model is presented as an extension of the standard neo-Wicksellian model. It explicitly incorporates banks, modeling the costly activities of deposit creation and loan provision. This introduces financial frictions, making monetary aggregates more influential in determining economic activity. The model also features bank deposits as a transaction medium, allowing for substitution with cash. Bank loans can influence aggregate supply and demand, depending on their modeling specifics, and alternative approaches are acknowledged. Government bonds serve a crucial role, held by banks and households for liquidity management. This introduces a spread between the CCAPM rate (from the consumption Euler equation) and the government bond rate, reflecting a liquidity premium. This spread helps explain the disconnect between money market rates and the rates influencing aggregate demand, and theoretically it buffers liquidity crunches from contractionary open market operations (OMOs). This has implications for interest rate surprises and the systematic component of the monetary policy interest rate rule. The model’s design incorporates the possibility of various impacts depending on the way bank loans are modeled, opening potential for future research into these aspects.

The NW model serves as a benchmark for comparison. It is described as a simplified version embedded within the BB model. The simplification involves removing banks and assuming that only cash is relevant for transactions. A money demand equation exists but is decoupled from the rest of the model, serving only to determine the open-market operations needed to implement the central bank’s interest rate rule. Notably, the NW model lacks financial frictions, aside from seigniorage taxes. The money demand equation could even be omitted for most applications. The authors emphasize that while simpler, the NW model's omission of banks and its treatment of monetary aggregates might lead to inaccuracies in capturing the intricacies of macroeconomic behavior. This sets the stage for the subsequent quantitative comparison and analysis of the impacts of these simplifications.

The core theoretical differences between the BB and NW models stem from the inclusion of banks and government bonds in the BB model. The BB model's financial frictions, arising from the costs of bank lending and deposit creation, make monetary aggregates inherently more significant. Government bonds, offering liquidity services to households and banks, create a liquidity premium and an endogenous spread between money market rates and the CCAPM rate. This spread directly impacts the transmission of monetary policy shocks and modifies the effects of OMOs. In contrast, the NW model's exclusion of banks and the associated liquidity premium results in a direct transmission of shocks, without the buffering effect. This theoretical difference is central to the paper's aim of evaluating the relative accuracy of the two models in representing macroeconomic behavior and informs the choice of comparative analytical techniques (impulse response functions, variance decompositions, and forecasting exercises) used in later sections. The contrasting treatment of monetary aggregates in each model also plays a pivotal role in the subsequent analysis of their respective predictive capabilities concerning inflation.

The calibration process for both the neo-Wicksellian (NW) and Banks and Bonds (BB) models is described. The models are calibrated to match quarterly averages of key macroeconomic variables. These variables include steady-state inflation and federal funds rates, as well as ratios representing the relative importance of various assets such as cash and government bonds in household and bank portfolios. Specific values are provided for parameters related to steady-state inflation and interest rates. The calibration also accounts for the distribution of labor across the banking and production sectors, ensuring that hours worked and wages are equal across both sectors in the steady state. This careful calibration sets the stage for the quantitative comparison between the two models, ensuring that the subsequent analysis is grounded in realistic parameter values. The calibration process ensures that the subsequent quantitative comparison is meaningful, providing a fair basis for comparing the models' responses to various economic shocks.

This section focuses on comparing the time series properties and impulse response functions generated by the two models. The comparison involves examining unconditional moments (second moments) and using variance decompositions to assess the relative importance of various shocks in explaining the behavior of key variables like inflation and output. The analysis of impulse response functions shows how unanticipated shocks (interest rate, productivity, government spending, and money demand shocks) propagate through each model. The results illustrate similarities and differences in the qualitative patterns, such as the transmission of interest rate shocks, and highlight any quantitative differences that arise. The ‘liquidity buffering effect’ of government bonds, unique to the BB model, is analyzed in the context of interest rate shocks. The section also explores the impact of productivity and government spending shocks, focusing on the quantitative differences in the models' responses. These quantitative comparisons are crucial for assessing the extent to which the simpler NW model adequately approximates the more complex BB model in terms of its implications for business cycles and monetary policy analysis.

A key aspect of the quantitative comparison is the evaluation of monetary aggregates' predictive power for inflation. The study uses artificial datasets generated from both models to estimate Vector Autoregressions (VARs). Bivariate VARs including inflation and the growth rate of different monetary aggregates are estimated, and additional VARs incorporate either the output gap or the nominal interest rate on government bonds to improve the predictive capacity of the models. The analysis focuses on the impulse response functions of inflation to shocks in the monetary aggregates. It investigates the predictive power of various monetary aggregates (M2, M2 velocity, dL, dLIAB, and cash balances) in forecasting inflation. The results are presented as median impulse response functions with confidence intervals to measure uncertainty. The findings reveal stark differences in the predictive power of monetary aggregates between the two models, with broader liquidity measures (dL, dLIAB) showing greater predictive ability in the BB model compared to the NW model. This reflects the importance of the features present only in the BB model.

The study concludes that the Banks and Bonds (BB) model provides a more comprehensive and arguably more accurate framework for analyzing monetary policy than the simpler neo-Wicksellian (NW) model. The quantitative differences observed between the models, particularly regarding the impact of government spending shocks on inflation, highlight the importance of considering the 'liquidity buffering effect' and the 'liquidity provision effect' inherent in the BB model but absent in the NW model. These effects, stemming from the inclusion of banks and government bonds in liquidity management, significantly influence the transmission of shocks and the overall macroeconomic dynamics. The findings suggest that the liquidity provided by government bonds plays a crucial role in dampening the impact of contractionary open market operations (OMOs), affecting the response to both interest rate shocks and other macroeconomic shocks. The more complete nature of the BB model provides a more accurate portrayal of these interactions.

The research underscores the differing roles of monetary aggregates in inflation forecasting, depending on the model used. The study finds that while simple monetary aggregates like nominal cash balance growth offer limited predictive power in both models, broader liquidity measures, including M2 velocity and total government liabilities (dLIAB), exhibit significant predictive power in the BB model. This difference emphasizes the importance of considering a more comprehensive financial structure when assessing the relationship between monetary variables and inflation. The greater predictive power of these broader measures within the BB framework is linked to its inclusion of features such as banks, government bonds used in liquidity management, and financial frictions. This highlights the need to move beyond simplistic representations of the monetary system when seeking to improve inflation forecasting accuracy. The findings suggest that in models with a more realistic treatment of the financial sector, broader liquidity aggregates might become better indicators for medium-term inflation trends.

The paper notes that interest rate rules violating the Taylor principle cause indeterminacy issues in the NW model, but not in the BB model. This is attributed to the non-Ricardian nature of the BB model, stemming from the liquidity services provided by government bonds. This finding has important implications for monetary policy analysis, demonstrating the need to consider more realistic model structures to address these technical issues. The authors propose areas for further research, including investigating the role of financial frictions and the bank loan mechanism in greater detail. This involves exploring alternative ways of modeling bank loan demand, potentially allowing the loan channel to have a more substantial effect on aggregate supply and demand. Also, incorporating model features to generate more persistence in the inflation process is suggested, to better distinguish between short-run and medium-run inflation dynamics. These extensions are expected to enhance the models' explanatory power and their ability to accurately capture the dynamics of real-world economies.