General Information & Client Services
  • Americas: +1.212.553.1653
  • Asia: +852.3551.3077
  • China: +86.10.6319.6580
  • EMEA: +44.20.7772.5454
  • Japan: +81.3.5408.4100
Media Relations
  • New York: +1.212.553.0376
  • London: +44.20.7772.5456
  • Hong Kong: +852.3758.1350
  • Tokyo: +813.5408.4110
  • Sydney: +61.2.9270.8141
  • Mexico City: +001.888.779.5833
  • Buenos Aires: +0800.666.3506
  • São Paulo: +0800.891.2518

In this article, I take a theoretical look at negative interest rates as a means to stimulate the economy. I identify key factors that may influence the volume of deposits held in the economy. I then empirically describe the unique situation of negative interest rates. I delve into the asymmetries that exist in the relationship between deposits and interest rates, considering Sweden as an example case.

As the global economy sags, with many regions in or teetering on the brink of recession, central banks are looking for new ways to boost economic activity and stave off deflation. The first approach attempted in the wake of the global financial crisis was quantitative easing, which arguably enjoyed some modest success in the US but which was viewed (fairly or unfairly) as a failure in various other important jurisdictions around the world.

More recently, a few countries have tried a new tack that involves charging banks for depositing reserve funds in the vaults of the central bank. Such policies have been implemented in Japan, which has been a pioneer in the use of radical monetary stimulation techniques; the eurozone; Switzerland; Denmark; and Sweden. The possibility that the US could chart a similar course cannot be easily dismissed.

On the asset side of the balance sheet, such a move is likely to have symmetric effects on commercial bank activity. With zero or razor-thin deposit rates, interest revenue earned by banks has been commensurately squeezed. One would expect this process to be further enhanced as monetary authorities push deeper into negative territory in a bid to boost their moribund economies.

The liabilities side provides an entirely different set of challenges. Banks want to maintain a consistent deposit base so they do not need to tap other more expensive sources of capital to fund their activities. They also want to carefully control the interest and noninterest costs of retaining these depositors, especially given the tight profit margins imposed from the asset side of the ledger. Bear in mind that many large banks are restricted from increasing the riskiness of their asset holdings due to the firm grip being applied by regulators in every advanced economy around the world.

Banks want to maintain a consistent deposit base so they can avoid using other, more expensive sources of capital to fund their activities.

Serious questions emerge regarding the effect of negative central bank deposit rates on the volume of deposits actually held. If, under negative rates, households and businesses continue to demand bank services to store their accumulated wealth in a risk-free manner, banks will be able to divert their attention to more profitable lending activities. Some, though, fear that negative rates will cause an exodus of depositors, forcing banks to raise capital from other places or to curtail their money-making operations. This paper seeks to cast some light on these issues from a macroeconomic perspective.

Building on Previous Research

Moody's Analytics has previously engaged in numerous studies of the effect of stressed economies on the aggregate deposit base. For example, Hughes1 looked at developing CCAR-style stress testing models for total US deposits in various categories. While the mix of deposits held in a range of products is acutely affected by macroeconomic effects, the overall level of funds held by banks tends to be only marginally impacted by generic macro stress. Recessions do cause growth to slow, but typically with quite a long lag. Lower interest rates, holding all else equal, tend to push clients away from CDs and term deposits in favor of more convenient forms of on-demand services. "Deposit recessions" – situations where overall volumes actually shrink – are very difficult to generate even under extreme macroeconomic duress.

"Deposit recessions," in which overall volumes actually shrink, are very rare even under extreme macroeconomic duress.

In a similar vein, Hughes and Poi2 extended this analysis to an individual institution with a long history of high-quality deposit data (North Carolina State Employees Credit Union). They found that account funds held by such an institution are likely to also grow in the midst of a recession, though pricing of services relative to the market plays an important role in determining the size of the slice eventually retained by the institution in question.

In a more pointed analysis, Poi, Malone, Hughes, and Zandi3 considered the effect of quantitative easing policies (and their subsequent reversal) on the deposit base held by US and Japanese banks. Using a variety of champion and challenger models for both jurisdictions, they found only marginal overall effects of the radical policies on the size of the deposit pie. The implication is that if central banks want to "push on the string" in a very low interest rate environment, or to remove said impetus from said ligature, it will have little overall effect on bank holdings of deposits.

The analysis in this paper should be viewed as an extension of earlier work.

We will begin by taking a theoretical look at the problem of negative rates and try to identify key factors that may influence the volume of deposits held in the economy. This discussion will then be used to guide an empirical study that will seek to tease out the asymmetries that exist in the relationship between deposits and interest rates. We will follow a similar approach to that used by Poi et al. and seek to identify the most useful test case for empirical analysis. Specifically, here we will consider the case of Sweden, mainly because the Riksbank was the first to employ negative rates as a key plank in its monetary policy back in 2009. Sweden is unique in that there were two distinct instances of negative deposit rates with a brief intervening period of positive rates. Poi et al. used Japan for the same reasons in their consideration of the impact of QE on overall deposit behavior.

Theoretical Analysis

In their economic activities, businesses and households generate stores of wealth. Part of this is then reinvested in risky ventures (like stocks and property) to generate additional future income streams. The dictates of a balanced portfolio and risk hedging then demand that part of the store of wealth be held in relatively riskless investment forms such as cash, government bonds, and insured bank deposits.

In a negative interest rate environment, it is safe to assume that the real economy is performing poorly. This implies both that the overall wealth generation engine is sputtering and that the range of available attractive, risky investment options is limited. The upshot of this, holding all else equal, is that investors will tend, at the margin, to retrench their risky investments but retain a high demand for the risk-free options. Though the overall wealth pie may be shrinking or stagnant, the riskless slice will tend to grow in scale throughout the period of real economic misfortune.

Figure 1. Effects of interest rates on various deposit types
Effects of interest rates on various deposit types
Source: Moody's Analytics

If interest rates on short-term government bonds and bank deposits are zero, in theory, people will be indifferent between holding cash, bonds, or deposits. If interest rates on bonds and deposits are negative, again in theory, people will retrench their holdings of these assets and seek to store their accumulated wealth in the form of cash. Pieces of paper depicting dead presidents always earn precisely zero percent annual interest regardless of any actions pursued by the central bank.

This simple analysis assumes that holding cash is as riskless as holding insured deposits.

Some media speculation has picked up on these theoretical musings, often with a comedic bent. The authors will opine about the likelihood of companies and/or households withdrawing all their funds and putting $100 bills in tin cans to be buried in the garden. Such activities carry a variety of risks that belie the notion that cash is a riskless way to store wealth. Mattresses can burn, treasure maps can be stolen, and buried notes can suffer water damage or be forgotten. On a more practical level, for households in the modern age, paychecks are electronically deposited and bills are paid automatically by remote computers. It is hard to imagine anyone with a measure of accumulated savings ever truly going off the banking grid.

Figure 2. Regression results: Demand deposits
Regression results: Demand deposits
Source: Moody's Analytics

For businesses and companies, the operational meshing with banks is even more extreme than it is for households.

Figure 3. Regression results: Small time and savings deposits
Regression results: Small time and savings deposits
Source: Moody's Analytics

Moreover, imagine for a moment that Apple, which reportedly holds more than $200 billion in cash reserves, decided to avoid paying the negative interest rate deposit charges imposed by their bankers and to instead bury 200 tons of $100 banknotes in the hills of Cupertino. Now imagine the meeting taking place between Apple's CFO and the company's army of external auditors.

It's just never going to happen.

Given that a range of unobserved risks and operational rigidities associated with holding cash remains, the question of the impact of negative rates on deposit holdings is squarely empirical in nature. We will consider other theoretical musings – notably the effect of the carry trade – in our empirical discussion of the Swedish economy.

Empirical Findings

For a variety of technical and policy reasons, and due to a severe recession, the overnight deposit rate in Sweden first entered negative territory in the summer of 2009. As GDP growth bounced, the rate was then lifted in late 2010, though this move was highly controversial at the time due to the severe recession that was continuing to rage across much of Europe. The doubters were then proved right as economic clouds once more engulfed Sweden. Renewed recession led to a resumption of the negative rate policy in the summer of 2014. This situation remains in place today.

The set of circumstances endured by the Swedes sets up an ideal test case for an assessment of negative rates. The most important feature is that we observe two substantial, distinct periods during which the external policy treatment was applied, as well as two separate baseline periods during which more normal operations were undertaken. Moreover, the controversy regarding the initial removal of the policy in 2010 implies that the action can be viewed as exogenously undertaken by monetary authorities. Data sourced from Statistics Sweden is of high quality, with a long history allowing precise modeling to proceed. Regression results used for the following analysis are included in the accompanying figures.

Figure 4. Regression results: Large time deposits
Regression results: Large time deposits
Source: Moody's Analytics

We identify various deposit categories by decomposing statistics for monetary aggregates. Sweden publishes M0 through M3 – four separate categories that progressively aggregate longer-duration forms of bank deposits. The M0 category is very narrowly defined, incorporating highly liquid forms of central bank deposits, banknotes, and coins. M1 adds in deposits available on demand; M2 adds small time deposits and other forms of savings accounts; and M3 adds large time deposits typically owned by corporations and wealthy individuals. We therefore interpret M1-M0 as "demand deposits," M2-M1 as "small time/savings deposits," and M3-M2 as "large time deposits."

We augment this data using information regarding interest rates at a variety of terms. We seek to control for longer end yield curve dynamics that will allow us to focus our attention on the specific effects of the overnight deposit rate. Importantly, we also consider external channels through which the Swedish economy may interact with other countries and regions. To do this, we include net exports of goods and services, the exchange rate of the krona with the euro, and a variety of prevailing European interest rates.

This set of variables allows us to consider the effect of the carry trade on the behavior of domestic deposits. If safe returns at home are elusive, one option available to depositors involves investing their funds in a foreign currency-denominated account in which positive interest will be paid. Such investments carry exchange rate risk, but these potential misfortunes are sometimes adequately compensated by the available interest rate differential. The carry trade is generally considered an important dynamic in Japan's battle with deflation during the early 2000s. In the current environment, where rates are low everywhere, we should expect the carry trade to be far less prominent. Nevertheless, we are happy to sacrifice a couple of degrees of freedom to control for its potential impact in the work presented here.

The final set of controls focuses on the real domestic economy. In a nutshell, we give Swedish households and businesses the option of consuming their deposits or of reinvesting their savings in more risky forms of investment. These macroeconomic factors enter our models with a lag to stave off any accusations of potential endogeneity.

We transform all control variables to ensure an absence of unit roots.

Our primary focus here is in assessing the symmetry of the observed relationships between short-term policy rates and various deposit categories. We create a dummy variable that is generally zero but switches to unity if negative deposit rates prevail within the Swedish monetary system at the time. The third variable of key interest is an interaction between the deposit rate and the dummy.

Negative interest rates seemingly cause a slight – but statistically significant – reduction in total deposits held.

If the marginal effect of a change in rates is the same on either side of the zero frontier, the parameter on the interaction term in our regression will be precisely zero. Statistically, therefore, we can test the hypothesis that this phenomenon prevails in the data by using a t-test on the estimated coefficient in the model. Similarly, the inclusion of the dummy variable allows us to consider the specific marginal effect involved with crossing the zero frontier. If the act of moving from positive to negative rates causes a change in deposit behavior we will observe a level shift in the deposit growth rate as this unfolds. The included figures display the key regression results.

Across the range of positive deposit rates, the behavior of the key marginal effects is in line with our prior expectations. All three deposit categories are sensitive to rate shifts such that rate increases tend to accelerate the growth rate of underlying deposits. Of the three categories, again as expected, demand deposits are the least sensitive to rate changes across the positive part of the number line.

As the zero frontier is hit, large time and demand deposits both shift lower, though the latter effect is not statistically significant at the 5% level. Small time/savings deposits, meanwhile, do not suffer a noticeable level shift.

For all time deposits, the intuitive effect of an increase in rates continues to hold on the negative side of the zero rate boundary. For large time deposits, abstracting from the presence of the level shift, the effect of rates is symmetric in the sense that marginal rate changes have the same deposit growth impact on either side of zero. Going from, say, -1% to -2% will have the same growth implications for large time deposits as moving from 2% to 1%. For small time deposits, meanwhile, the rates effect is more pronounced on the negative side, implying that small time depositors become hypersensitive to rate cuts that make such holdings commensurately more expensive.

Things get really interesting when we consider demand deposits. After noting the slight downshift in such instruments when the zero mark is crossed, we can further observe demand deposit volumes tending to increase with further short-term rate cuts. Because demand deposits represent around 80% of all Swedish holdings, it is this effect that is the most economically significant of the findings across the three separate categories.

As rates fall further below zero, total deposits held by banks actually increase!

This is the most interesting finding of the empirical research. Though it is true that negative rates sap the performance of term deposits, the funds repatriated from this process do not do so in the form of cash. Rather, they exit in the form of demand deposits.

For banks, there are two implications of these findings. One is that crossing the frontier seemingly causes a slight – but statistically significant – reduction in total deposits held. For Sweden, this initial effect amounted to around 4% of the total deposit base. As further rate cuts are effected, though, part of this reduction can be clawed back in the form of perhaps higher than anticipated growth in demand deposits. The second implication is that the duration of the deposit book will tend to decline as funds are moved out of term deposits and CDs into vehicles that allow funds to be drawn instantly.


Radical central bank policy changes are rarely welcomed by bankers. With the global economy still trying to shake off the lingering effects of the Great Recession more than seven years post-Lehman Brothers, it is little wonder that monetary authorities are seeking new directions in their stimulative efforts. Recently, this push has moved in the direction of punishing savers for holding risk-free investment forms. Such an unprecedented move has banks worried for the safety of their deposit holdings.

In a series of papers, Moody's Analytics has explored the empirical effects of radical policy shifts on deposits. In general, it is observed that such moves often have a significant impact on the form of deposits but not an especially large effect on their scale. The findings of this paper are fully consistent with these observations. Though negative rates will cause term holdings to shrink, this effect will be more than offset by a rise in demand deposits.


1 Hughes, Anthony, Deposit Stress Testing, Moody's Analytics whitepaper, June 2013.

2 Hughes & Poi, Improved Deposit Modeling: Using the Moody's Analytics Pre-Provision Net Revenue Factors Library to Augment Internal Data, Moody's Analytics whitepaper, July 2015.

3 Hughes, Malone, Poi, & Zandi, Quantitative Easing and Bank Deposits, Moody's Analytics whitepaper, October 2015.

As Published In:
Related Insights

Auto Finance Insights: How Vehicle Information Informs Credit Risk Measures

Tony Hughes and Michael Vogan share valuable insights for managing your auto lending business more effectively.

April 2018 WebPage Dr. Tony HughesMichael Vogan

Market Share-Based Credit Analysis

A counterpoint to traditional credit performance analyses.

March 28, 2018 Pdf Dr. Tony Hughes

The Data Revolution: Gaining Insight from Big and Small Data

In this article, we explore the importance of small data in risk modeling and other applications and explain how the analysis of small data can help make big data analytics more useful.

December 2017 WebPage Dr. Tony Hughes

The Effect of Ride-Sharing on the Auto Industry

Many in the auto industry are concerned about the impact of ride-sharing. In this article analyze the impact of ride-share services like Uber and Lyft on the private transportation market.

July 2017 Pdf Dr. Tony Hughes

The Effect of Ride-Sharing on the Auto Industry

In this article, we consider some possible long-term ramifications of ride-sharing for the broader auto indust

July 2017 WebPage Dr. Tony Hughes

"How Will the Increase in Off-Lease Volume Affect Used Car Residuals?" Presentation Slides

Increases in auto lease volumes are nothing new, yet the industry is rife with fear that used car prices are about to collapse. In this talk, we will explore the dynamics behind the trends and the speculation. The abundance of vehicles in the US that are older than 10 years will soon need to be replaced, and together with continuing demand from ex-lessees, this demand will ensure that prices remain supported under baseline macroeconomic conditions.

February 2017 Pdf Dr. Tony HughesMichael Vogan

How Will the Increase in Off-Lease Volume Affect Used Car Residuals?

Increases in auto lease volumes are nothing new, yet the industry is rife with fear that used car prices are about to collapse. In this talk, we will explore the dynamics behind the trends and the speculation.

February 2017 WebPage Dr. Tony HughesMichael Vogan

Economic Forecasting & Stress Testing Residual Vehicle Values

To effectively manage risk in your auto portfolios, you need to account for future economic conditions. Relying on models that do not fully account for cyclical economic factors and include subjective overlay, may produce inaccurate, inconsistent or biased estimates of residual values.

December 2016 WebPage Dr. Tony Hughes

The Value of Granular Risk Rating Models for CECL

Granular risk rating models allow creditors to understand the credit risk of individual loans in a portfolio, facilitating underwriting and monitoring activities. In this webinar we will outline the value of granular risk rating models for CECL.

November 2016 WebPage Christian HenkelDr. Tony Hughes

Improved Deposit Modeling: Using Moody's Analytics Forecasts of Bank Financial

In this article we demonstrate how to combine our forecasts of bank financial statements with internal data to produce forecasts that better reflect the macroeconomic environment posited under the various Comprehensive Capital Analysis and Review scenarios.

August 2016 Pdf Dr. Tony HughesBrian Poi

AutoCycle™: Residual Risk Management and Lease Pricing at the VIN Level

We demonstrate the core capabilities of our vehicle residual forecasting model to capture aging and usage effects and illustrate the material implications for car valuation of different macroeconomic scenarios such as recessions and oil price spikes.

May 2016 Pdf Dr. Tony Hughes

Benefits & Applications: AutoCycle - Vehicle Residual Value Forecasting Solution

With auto leasing close to record highs, the need for accurate and transparent used-car price forecasts is paramount. Concerns about the effect of off-lease volume on prices have recently peaked, and those exposed to risks associated with vehicle valuations are seeking new forms of intelligence. With these forces in mind, Moody's Analytics AutoCycle™ has been developed to address these evolving market dynamics.

May 2016 Pdf Dr. Tony HughesDr. Samuel W. MaloneMichael Vogan, Michael Brisson

Alternatives to Long-Term Car Loans?

In this article, our experts focus on two recent developments: how to manage lease-term or model-year concentration risk and how to find affordable finance options for subprime or near-prime sector.

February 2016 Pdf Dr. Tony Hughes

Small Samples and the Overuse of Hypothesis Tests

With powerful computers and statistical packages, modelers can now run an enormous number of tests effortlessly. But should they? This article discusses how bank risk modelers should approach statistical testing when faced with tiny data sets.

December 2015 WebPage Dr. Tony Hughes

Do Banks Need Third-Party Models?

This article discusses the role of third-party data and analytics in the stress testing process. Beyond the simple argument that more eyes are better, we outline why some stress testing activities should definitely be conducted by third parties.

December 11, 2015 WebPage Dr. Douglas DwyerDr. Tony Hughes

Stress Testing Used-Car Prices

In this presentation we presented a quantitative methodology for incorporating economic factors into car price forecasts.

August 2015 WebPage Dr. Tony HughesMichael Vogan

Systemic Risk Monitor 1.0: A Network Approach

In this article, we introduce a new risk management tool focused on network connectivity between financial institutions.


Forecasts and Stress Scenarios of Used-Car Prices

The market for new cars is growing strongly and lessors need forecasts and associated stress scenarios of future vehicle value to set the initial terms, to monitor the performance of their book and to stress-test cash flows. This presentation offers insight and tools to help lessors in this pursuit.

May 2015 Pdf Dr. Tony Hughes, Zhou Liu, Pedro Castro

Measuring Systemic Risk in the Southeast Asian Financial System

This article looks back at the Asian financial crisis of 1997-1998 and applies new methods of measuring systemic risk and pinpointing weaknesses, which can be used by today’s financial institutions and regulators.


Multicollinearity and Stress Testing

Multicollinearity, the phenomenon in which the regressors of a model are correlated with each other, apparently causes a lot of confusion among practitioners and users of stress testing models. This article seeks to dispel this confusion.

May 2015 WebPage Dr. Tony HughesBrian Poi

What if PPNR Research Proves Fruitless?

This article addresses how banks should look to sources of high-quality, industry-level data to ensure that their PPNR modeling is not only reliable and effective, but also better informs their risk management decisions.

May 2015 WebPage Dr. Tony Hughes

Vehicle Equity and Long-Term Car Loans

In this article, we consider the increasing prevalence of long term loans and use the AutoCycle™ wholesale price forecasts to uncover equity held by the borrower under different economic scenarios.

April 2015 Pdf Dr. Tony Hughes

Modeling the Entire Balance Sheet of a Bank

This article explores the interaction between a bank’s various models and how they may be built into a comprehensive stress testing framework, contributing to the overall performance of a bank.

November 2013 WebPage Dr. Tony Hughes

Is Now the Time for Tough Stress Tests?

The banking industry needs a regulatory framework that is carefully designed to maximize economic outcomes, both in terms of stability and growth, rather than one dictated by past banking sector excesses.

November 2013 WebPage Dr. Tony Hughes

Stressed EDF Credit Measures for Western Europe

In this paper we describe the modeling methodology behind Moody's Analytics Stressed EDF measures for Western Europe. Stressed EDF measures are one-year, default probabilities conditioned on holistic economic scenarios developed in a large-scale,structural macroeconometric model framework.

October 2012 Pdf Danielle Ferry, Dr. Tony Hughes, Min Ding

Stressed EDF™ Credit Measures for North America

In this paper we describe the modeling methodology behind Moody's Analytics Stressed EDF measures. Stressed EDF measures are one-year, default probabilities conditioned on holistic economic scenarios developed in a large-scale, structural macroeconometric model framework. This approach has several advantages over other methods, especially in the context of stress testing. Stress tests or scenario analyses based on macroeconomic drivers lend themselves to highly intuitive interpretation accessible to wide audiences – investors, economists, regulators, the general public, to name a few.

May 2012 Pdf Danielle Ferry, Dr. Tony Hughes, Min Ding

The Moody's CreditCycle Approach to Loan Loss Modeling

This whitepaper goes in-depth into the Moody's CreditCycle approach to loan loss modeling.

Previewing This Year's Stress Tests Using the Bank Call Report Forecasts

To capture a bank’s real capacity to withstand an adverse economic scenario, the best approach is to start with forecasts produced only for accuracy and then apply a conservative overlay as regulation requires. Such forecasts capture mitigating forces such as flight to safety.


Stress Testing and Strategic Planning Using Peer Analysis

Banks face the difficult task of building hundreds of forecasting models that disentangle macroeconomic effects from bank-specific decisions. We propose an approach based on consistently reported industry data that simplifies the modeler’s task and at the same time increases forecast accuracy.

Are $10 Billion-to-$50 Billion Banks Overregulated?

Moody’s Analytics tests whether relaxing capital standards for $10 billion-to-$50 billion banks will boost banking and encourage greater lending.

WebPage Dr. Tony Hughes