Bank Funding Dynamics Between Retail Deposits and Wholesale Funds: Implications for Regulations

Wholesale funding sources primarily consist of short-term funds or deposits provided by foreign institutions, nonfinancial firms, non-depository financial institutions and state or local municipalities.

We leave the discussion of sample period in detail to Section 4.

We leave the detailed theoretical discussion in the hypothesis development section.

In this analysis, we utilize a more comprehensive liquidity ratio that incorporates Treasury and agency debt: broader liquidity ratio. Variable construction details are shown in Table A1.

Although our bank-level panel regression analysis partially mitigates the omitted variable bias concern that exists in the time-series analysis at the aggregate level, there are still endogeneity concerns that need to be addressed. On one hand, although we control for lending growth in our model specifications, loan demand or lending opportunities may still affect the growth of retail deposits and wholesale funds simultaneously because they jointly shape the bank lending process. On the other hand, characteristics of the suppliers of wholesale funding (i.e., wholesale funding investors) may confound this substitution effect between the two funding sources because they may or may not be informationally sensitive to curbing their funding supplies to banks (Perignon et al. 2018).

In terms of transparency, we rely on the public status of BHCs in the sample as prior literature indicates that public banks are more transparent than private ones. We leave the discussion of this in detail to Section 5.10.2.

Wholesale funds are heterogeneous as they vary by riskiness and maturity. In this paper, we categorize them into time deposits of $100,000 or more, foreign deposits, subordinated notes and debentures, Federal funds purchased and securities sold under agreements to repurchase, and other borrowed money.

A counterargument to this logic is that the effective maturities of retail deposits are similar to CDs given depositor behavior; however, CDs are relatively less liquid than short-term funding sources such as Federal funds so our expectation holds.

The data can be retrieved from the FDIC’s website: https://​www.​fdic.​gov/​bank/​analytical/​qbp/​.

FR Y-9C report data source: https://​www.​chicagofed.​org/​applications/​bhc/​bhc-home.

Removing assets growth or lending growth above a certain threshold (i.e., 10%) has been commonly used in the banking literature (e.g., Gatev and Strahan 2006; Acharya and Mora 2015; Chen et al. 2022).

Table A1 provides detailed descriptions of the aggregate time-series variables, bank-level characteristics and other relevant micro variables that are used in this study.

According to Acharya and Mora (2015), wholesale funds are the sum of large time deposits, deposits held by foreign institutions, subordinated debt and debentures, gross Federal funds purchased, repos and other borrowed money. Choi and Choi (2021) define wholesale funds as the sum of wholesale deposits, Federal funds and repos, and other borrowed money. Lin (2020) defines wholesale funds as the sum of brokered deposits, foreign institutions’ deposits, other borrowed money and Federal funds purchased.

Branch offices and their corresponding parent BHCs are linked based on the variable, “rssdhcr”, in the Summary of Deposits database.

We also plot the figures by quarter. The unreported results show that in the years after 2012, we notice a mechanical relationship between the growth rates of retail deposits and wholesale funds. In the second quarter of each year, the median and average growth rates of wholesale funds increase to levels higher than those of retail deposits while, at the same time, the growth rates of retail deposits sharply decrease.

We also employ a broader definition of wholesale funds that includes brokered deposits and these results are found in Table A2. Nonetheless, the results indicate that our bank-level findings remain unchanged.

Beginning in 2017, the FR Y-9C report increased the cutoff amount from $100,000 to $250,000 on time deposits. Because of this inconsistency in the financial report, we choose to shorten the sample period to end in 2016. However, the actual deposit amounts between $100,000 and $250,000 are likely not significant. In order to verify, we conduct an additional analysis using the cutoff of $100,000 for the data before 2017 and $250,000 for the data after 2016. The corresponding items for these two cutoffs are BHCB2604 (total time deposits of $100,000 or more) and BHCBJ474 (time deposits of more than $250,000). By combining these two cutoffs in two separate periods in the sample, we can create a comprehensive sample that extends to the first quarter of 2020 right before the start of the global outbreak of COVID-19. Hence, we conduct a corresponding analysis for our bank-level baseline regressions as a robustness check. Nevertheless, our results remain unchanged and are displayed in Table A8.

As the substitution effect could be directional for both retail deposits and wholesale funds, we also use a seemingly unrelated regression model to test the dynamics in these directions. The untabulated results show that our main finding here continues to hold.

We interpret and display the results of Table 5 according to the order in Hypothesis 2.

However, subordinated notes and debentures are relatively easy to issue compared to other types of wholesale funding, which may influence this result.

These subfigures plot the wholesale funds ratios of large and small banks from 2002 to 2016. Figure 4a and b display the mean and median ratios, respectively, of banks using a $10 billion threshold to classify bank size. Figure 4c and d display the mean and median ratios, respectively, of banks using a $50 billion threshold to classify bank size. Moreover, Fig. 4e through h plot the retail deposits ratios for large and small banks, which exhibit opposite patterns compared to the corresponding figures with wholesale funds ratios.

For details regarding U.S. stress tests, see this link: https://​www.​fdic.​gov/​regulations/​laws/​rules/​5000-5360.​html#:​~:​text=​Under%20​the%20​Stress%20​Test%20​Rule%2C%20​each%20​covered%20​bank%20​is%20​required,all%20​categories%20​of%20​financial%20​data.

An anonymous reviewer expressed concerns that we gave more weight to smaller banks in early years when creating a constant sample. Thus, as a robustness check, the reviewer suggested that we adopt the strictest reporting size threshold adjusted by inflation, which is a reasonable alternative approach. To that end, we use the threshold of $3 billion that was effective in 2018 and adjust this backwards by inflation throughout our sample for periods prior to 2018 using the series “inflation, consumer prices for the United States” obtained from the Federal Reserve Bank of St. Louis. We then reconduct most of the bank-level baseline regressions that are found in Tables 3 through 6. Using this threshold considerably reduces the sample observations to fewer than 10,000. Nonetheless, the modified regression results remain unchanged by adopting this alternative approach except in Panel B of Table 4, where we investigate the asymmetry of the substitutional relationship between retail deposits and wholesale funds by using subsamples of only negative or positive retail deposits growth. We attribute this inconsistency to small samples as this approach further significantly decreases observations. All results from this alternative approach are displayed in Tables A4 through A7, accordingly.

To make this figure more intuitive, we take the absolute values of these coefficients over the sample period when we draw the figure.

We also repeat the regression from Column (4) of Table 9 by year throughout the sample period and these results can be found in Table A3.

Due to valid economic reasons, banks might not want to maintain current levels of lending during a financial crisis. However, our results show that banks with stronger active management of wholesale funds for retail deposits were not better suppliers of loans during this funding shock.

On the other hand, all else being equal, a bank’s access to wholesale funding markets is a crucial component of its business and without such access a bank would be considered very risky. In theory, banks should seek to strike a balance between retail deposits and wholesale funds without overrelying on the latter.

These interaction terms are broader liquidity ratio and high liquidity dummy which identifies banks with broader liquidity ratios higher than the median.

While this approach mitigates this confounding effect, it is impossible to consider all of the factors that may affect a bank’s decision to lend as the lending decision is influenced by many factors including the availability of retail funds and economic conditions. These factors may, in turn, affect a bank’s reliance on wholesale funding and thus the substitution effect.

We utilize a dummy variable because public banks are subject to more rigorous regulations and are relatively more transparent than private ones in financial disclosure.

Public companies typically have better financial disclosure quality and are more transparent than private ones.

Overall, our study is not without limitations as the dynamics between retail deposit and wholesale fund flows are quite complex and not truly exogenous. The factors driving shifts in retail deposits influence the wholesale funding markets and vice versa. Furthermore, changes in banks’ lending opportunities due to market conditions further complicate the relationship.