Interest-rate and calendar-time effects in money market fund and bank deposit cash flows

In addition to the calendar-based preferred habitat literature, Simon (1994) shows that T-bill investors will not trade into adjacent bills to pick up additional yield.

Deposits are the main source of funds for banks and the key to their profitability. Based on FDIC data as of December 31, 2005, domestic deposits financed 57 percent of the $9.04 trillion of total assets in the banking industry and 97% percent of the net loans and leases made by commercial banks. The domestic deposits can be categorized as transaction accounts (14.5% of total deposits), money market deposit accounts (MMDAs) (39.1%), savings deposits excluding MMDAs (14.4%), small time deposits (14.6%), and large time deposits (17.4%). The double digit percentages in each category indicate how important each source is to the banking industry.

Aggregate balances are reported each Wednesday, while T-bill yields are collected on Mondays. We note that the end of the sample pre-dates the financial crisis.

While all of the weekly balance series are nonstationary (which would be expected), the null of non-stationarity is rejected for all of the dependent variables (weekly percentage changes in the balance) at the 1% level of significance or better.

According to the 2006 Investment Company Fact Book, the average maturity for MMFs was 45.4 days from 1983 to 2005 and varied from an annual average of 28 to 58 days with 11 of the years below and 13 of the years above the average of 45.4 days. Using the 45.4 day average maturity and a naïve assumption that an equal amount of securities is purchased each day, the average returns for the past 91 days (13 weeks) should be a good proxy for the amortized-cost returns on those MMFs for the next week. We have run different specifications of the model (e.g., with the dependent variable specified as the level or natural logarithm of the aggregate balance, and the key independent variable specified as a relative spread between current 13-week T-bill yield and the average of 13-week T-bill yields over the past 13 weeks or six weeks). The results are qualitatively similar.

Corporate cash disbursements usually flow through a central checking account for audit purposes. Individuals tend to function in a similar manner. Even when checks are written directly on MMF balances, payees will initially deposit them into their checking accounts, and the pattern of falling MMF balances and increasing demand deposit balances before year- and quarter-ends would be preserved.

Farinella and Koch (2000) also test for tax dates other than April 15 in their study of MMF cash flow patterns. We do not find significant balance changes around those dates (the results are not reported).

We find no evidence of structural breaks. Accordingly, we estimate all regressions across the entire sample period. Additionally, we note that the ARCH and GARCH parameter estimates in the variance equation are statistically significant in all estimation of Eq. 1. We do not report the variance equation estimates in the interest of brevity. They are available upon request.

One can argue that investors may store funds in MMFs to later invest in the stock market. If this is true, MMF balance changes may be negatively correlated with stock index returns. However, weekly money market fund flows are not significantly correlated with contemporaneous and lagged stock market returns (these results are not reported).

Table 1 shows a mean balance in institutional funds of $456.7 billion, so a 1.58% change is $7.2 billion.

Table 1 shows a mean balance of $496.3 billion, so the change of 0.19% is $943 million.

During our sample period, the average portfolio maturity of an MMF is 45 days. This means a significant portion of an MMF portfolio matures and converts to cash each day. This allows MMF managers to cover net cash outflows through security maturities instead of securities sales.

Because we use weekly data, the first weekly change in a year (quarter) may be dominated by the cash flows that occurred between the last available observation in the year (quarter) and the end of the year (quarter). The weekly data points are Wednesdays.

Hagerman (2007) notes that institutional MMFs hold 27% of corporate liquidity and are the number one investment choice of corporate treasurers. Ogden (1987) suggests that calendar year-end effect in money markets is strong, in part, because of the concentration of corporate cash obligations at the year-end.

To get a feel for the magnitude of cash flows, we also ran Eq. 1 with changes in billions of dollar as a dependent variable. The year-end effect results in about $24 billion flowing out of MMFs with about $16.5 billion of these flows from institutional MMFs. Additionally, we observe about $6 billion leaving MMFs at tax time. The results are not reported but available upon request.

This series is only available starting on December 31, 1993, thus not fully matching our sample period. The details of constructing this variable can be found at http://​research.​stlouisfed.​org/​publications/​net/​NETJan2010Append​ix.​pdf.

While the flow of funds depends to some extent on the perceived riskiness of the investments, both bank deposits and MMFs were perceived to have extremely low levels of risk during our sample period. MMF and deposit balances are not sensitive to changes in the credit spread (BAA—AAA).