The Unpredictable Nature of Demand Deposits: Why Consumer Behavior Matters in Balance Sheet Management

For ALCO teams at Community Banks and Credit Unions across the U.S., demand deposits have always felt like the bedrock of balance sheet management. But in today’s digital-first, sentiment-driven environment, that bedrock is shifting. What used to be predictable is now anything but.

When “Safe” Isn’t Really Safe

Ask any banker under $5bn in assets what keeps their NIM alive, and they’ll likely give you the same answer: demand deposits. They’re cheap, they stabilize earnings, and they’ve always been assumed to stick around.
But assumptions can be dangerous. In recent years, deposit outflows have shown how quickly loyalty can evaporate when customers feel spooked, tempted, or simply bored with low yields. If your ALCO is still running with the idea that deposits are “stable by nature,” it’s time to think again.

Why Demand Deposits Matter So Much

  • The Funding Advantage

    Demand deposits lower the weighted average cost of funds and reduce reliance on wholesale borrowing. For community banks and credit unions, that funding advantage can mean survival in a competitive market.

  • The Balance Sheet Anchor

    In theory, these deposits act like an anchor: reliable, slow to move, and providing stability. They’re the invisible hand keeping your interest rate risk in check.

  • The Hidden Risk

    But anchors can drag. When outflows surge, the impact isn’t linear. It can create liquidity crunches, distort ALM reports, and force last-minute funding moves that erode profitability.

Case Study 1: The March 2023 Stress Event

In March 2023, U.S. banks of all sizes saw sudden deposit outflows. But community banks were hit hardest.

Why?

  • Less diversified deposit bases.

  • Heavier reliance on local businesses and niche communities.

  • Social media fueling panic faster than management could respond.

What Happened?

Some institutions saw outflows equivalent to 10–20% of total deposits in a matter of days. Stress testing models that assumed slow runoff were blindsided.

The Lesson for ALCO:

Don’t model outflows as a trickle. In a digital world, they can turn into a flood overnight.

The Behavioral Challenge: Customers Don’t Behave Like Models Assume

Traditional Assumptions

For decades, ALM models treated demand deposits as semi-permanent. Decay rates were set once, based on local history, and left untouched for years.

Today’s Reality

Behavioral drivers now include:

  • Digital ease: transfers happen in seconds, not days.

  • Rate competition: customers chase yield aggressively when rates rise.

  • Macro sentiment: fear of inflation or recession drives “safety-first” moves.

  • Generational shifts: Gen Z customers don’t see banking as a “relationship.” For them, it’s just another app.

Regulatory Pressure

Basel III, IFRS 9, and IRRBB all emphasize behavioral justification. FDIC and OCC have issued guidance noting that “legacy decay rates are insufficient in a digital environment.” Translation: prove your assumptions, or prepare for scrutiny.

Case Study 2: A Rural Credit Union’s Wake-Up Call

A Midwest credit union with $1.2bn in assets assumed long-term loyalty from its member base. Historically, runoff rates were below 2%.

In 2024, as inflation drove up rates, younger members moved funds into online investment apps. Outflows hit 12% in a single quarter.

The credit union had to scramble for liquidity via the Federal Home Loan Bank, paying rates that crushed their margin.

Moral of the story (Oh Yes, there is always a Moral): even in close-knit communities, loyalty doesn’t guarantee stability.

Analytical Approaches: Raising the Bar

1. Statistical Models

Regression and time-series models can still provide useful baselines. But they must account for seasonality (tax season, holiday spending) and cyclicality (rate cycles, recessions).

2. AI and Machine Learning

Modern approaches include:

  • Survival analysis models to estimate “time to runoff.”

  • LSTM neural networks for detecting nonlinear patterns in transaction flows.

  • Ornstein–Uhlenbeck and Lévy processes to model mean reversion with rare shocks.

Even small banks can access cloud-based AI tools that scale without heavy infrastructure.

3. Stress Testing

Run scenarios beyond the “mild stress” case. Include extreme but plausible outflows,  because history shows they happen.

4. Segmentation

Retail vs. corporate, Boomers vs. Gen Z, rural vs. urban — deposits are not one-size-fits-all. Segmentation prevents distorted results.

The ALCO Implication: Errors That Cost Millions

When assumptions break, the ripple effects can be brutal:

  • Liquidity Coverage Ratio (LCR) appears stronger than reality.

  • Interest rate gap reports give a false sense of protection.

  • Capital planning turns overly optimistic or excessively conservative.

At the end of the day, ALCO’s role is to keep the balance between optimism and prudence. That balance is impossible without realistic deposit modeling

Expanding the Regulatory Lens

  • Basel III & IRRBB

    Require banks to justify behavioral assumptions with quantitative evidence.

  • FDIC Guidance (2020 Community Banking Study)

    Warns that non-maturing deposits are “volatile under digital stress.”

  • OCC Expectations

    OCC examiners expect banks to document deposit betas and stress test outcomes.

  • NCUA Interagency Guidance

    Specifically calls on credit unions to align funding strategies with contingency plans, not historical averages.

Takeaway: every regulator is saying the same thing, show your math.

A Forward-Looking Playbook

For community banks, the future of deposit management lies in combining human judgment with machine-driven insights.

  • Scenario-Based Forecasting

    Don’t just run “base case” models. Build stress cases for inflation spikes, unemployment shocks, and social media-driven runs.

  • Real-Time Monitoring

    Use transaction-level analytics to catch unusual flows. Think of it as an early warning system.

  • Generational Segmentation

    Boomers may still value loyalty. Millennials and Gen Z chase integration and convenience. Treat them differently.

  • Contingency Plans

 

    Maintain access to:

  • Discount Window,
  • FHLB advances,
  • Brokered deposits (as a last resort).

Practical Tools for ALCO

1. Behavioral scoring models by depositor.

2. Heat maps of deposit concentrations.

3. Deposit betas tracking dashboards.

4. Liquidity early-warning indicators.

5. Digital channel monitoring (ACH, Zelle, Venmo activity).

The Future: AI as a Game-Changer

  1. Review and update deposit decay assumptions.
  2. Segment deposits by type, generation, and geography.
  3. Run stress tests with extreme but plausible outflow scenarios.
  4. Track deposit betas against Fed rate moves.
  5. Build a liquidity early-warning dashboard.
  6. Establish real-time monitoring of digital flows.
  7. Document regulatory justifications for decay rates.
  8. Strengthen contingency funding options.
  9. Educate the board on behavioral risks.
  10. Treat deposits as dynamic, not static, in every ALCO discussion.

Stickiness Isn’t What It Used to Be

Demand deposits remain the cheapest and most strategic source of funding for community banks and credit unions. But “cheap” does not mean “safe.”
The only predictable thing about deposits today is their unpredictability. For ALCO, that means shifting from static assumptions to dynamic, data-driven strategies that combine traditional judgment with AI-enabled insights.
Because when confidence wavers, deposits don’t wait. Neither should you.

References

Regulatory & Supervisory Guidance

Basel Committee on Banking Supervision – Interest Rate Risk in the Banking Book (IRRBB)

IFRS 9 – Financial Instruments

FDIC – 2020 Community Banking Study

OCC – Liquidity Risk: Comptroller’s Handbook

NCUA – Interagency Policy Statement on Funding and Liquidity Risk Management (2010)

Federal Reserve – Supervision and Regulation Letter SR 10-6: Interagency Policy Statement on Funding and Liquidity Risk Management

Research & Academic Papers

SSRN – Modeling Non-Maturing Demand Deposits

ArXiv – Ornstein–Uhlenbeck Processes with Lévy Noise in Financial Modeling

FRB Richmond – Bank Run Interrupted: Modeling Deposit Withdrawals with Generative AI

ResearchGate – Modeling Non-Maturing Deposits: Threshold of Separation Between Volatile and Stable Volumes

Industry & Think-Tank Reports

Cornerstone Advisors – Deposit Displacement Is Killing Banks and Credit Unions

CSBS – 2023 Community Bank Survey

Community Banking Connections – Community Bank Liquidity: A Post-Pandemic Perspective

Community Banking Connections – The Digital Deposit Challenge

 

Share the Post:

Related Posts

Find out more.

Fill out the form with your details and we will send the material to your email

Find out more.

Fill out the form with your details and we will send the material to your email

Find out more.

Fill out the form with your details and we will send the material to your email

Find out more.

Fill out the form with your details and we will send the material to your email

Find out more.

Fill out the form with your details and we will send the material to your email

Find out more.

Fill out the form with your details and we will send the material to your email