Money funds were already government-heavy, and the government share jumped +2.0 points in the one month the March 2023 bank stress fell in

What we anchor to

Event time is measured in months from the March 2023 N-MFP report month, the event's t equals zero. Silicon Valley Bank was closed by the Federal Deposit Insurance Corporation on 10 March 2023 and Signature Bank on 12 March, so the deposit run and the flight into money funds land inside the March 2023 month-end filings. The window runs six months each side, September 2022 to September 2023, which sits entirely inside the panel and entirely before a later classification change, so every movement in it is a movement of money rather than of definitions.

One piece of context predates the panel and this note therefore cannot show it. The 2016 SEC money-market-fund reform, the floating net asset value and the liquidity fees and gates placed on institutional prime funds, drove the large one-time rotation out of prime and into government funds. That happened in 2016, six years before these filings begin, so the panel opens with the complex already government-dominated rather than showing the reform itself. The site's short-term funding page carries the OFR series that reaches back to 2013 and does show that earlier rotation.

What we measure, and how it is counted

• One series, shared with the money-fund module. The line is the government share of total money-market-fund net assets, built from the same Form N-MFP category series the site's money-market-fund page plots, read here through the same query rather than re-fetched. The three N-MFP fund categories (government, prime, tax-exempt) partition the complex, so each month's total and each category's share are the reused series added up, not a separate pull.
• A share of a total, not a dollar figure of safety. The headline series is the government funds' slice of aggregate net assets, in percent. The prime share is very nearly its mirror, since the two plus a roughly two-point tax-exempt sliver sum to the whole, so the table below carries both columns while the chart plots the government share alone. Feeder funds are excluded throughout, exactly as the SEC's own report excludes them.
• Event offsets are calendar months, never row steps. Each month is placed relative to the event by its calendar position (year and month), not by counting a fixed number of rows back or forward. The N-MFP panel is gapless, but a row-offset lag or lead over a monthly series is a defect this site has hit before and now anchors every window to the calendar to avoid.
• Jumps use the displayed numbers. Shares are rounded to a tenth of a point and net assets to a hundredth of a trillion, as shown, and every change on the page is the difference of those rounded values, so the arithmetic on the page matches the arithmetic behind it.

The government share, on one clock

Read the six months before the event first. The split had been moving the other way: the government share fell -3.0 points from September 2022 to the month before the stress, and the prime share rose +2.9 points over the same stretch to its panel high of 22.0% in February 2023. Then, in the March 2023 report month, that drift reversed in one step: the government share jumped +2.0 points to 77.8% while the prime share fell -1.8 points. Over the following six months the spike partly unwound, the government share easing -0.5 points by September 2023 as the prime share recovered +0.5, and the prime share went on climbing back to 21.9% by March 2024, just short of its pre-stress high, before a later classification change moved the categories.

The window, month by month

At the event month, government funds took in $0.39T while prime funds gave up $0.02T and the complex as a whole grew $0.36T, so the new cash and the money leaving prime both landed in government funds. Each change is the difference of the displayed rounded values.

Is the jump really the event, or just noise?

The March 2023 move is the largest genuine one-month change in the government share anywhere in the panel: +2.0 points, larger than every other month's month-to-month change. The one still larger reading, +3.3 points at June 2024, is not a flow at all: it is the N-MFP form revision of June 2024, which re-labeled how government and Treasury funds are classified and mechanically lifted the reported government share. This note excludes that classification month from the comparison and keeps the event window well before it, so the March 2023 figure stands as the sharpest movement of actual money in the sample.

What this cannot tell you

• This is timing, not cause. The note records when the government share rose and when it eased around the March 2023 report month. It does not identify why cash moved. The month coincided with two large bank failures, uninsured-deposit outflows, and a widely reported flight into government money funds, all documented context, but this note attributes the movement of the share to none of them and tests no mechanism.
• A short panel around one episode. There is exactly one bank-stress episode in these 49 months, so the note describes a single event window rather than an average over many. It shows what the government and prime shares did around March 2023; it cannot say whether money funds behave this way around bank stress in general.
• A category is not a measure of safety. Government and prime are N-MFP fund categories defined by what a fund may hold, not a ranking of risk. A rising government share is a shift in where cash sits, not evidence that the complex became safer or that any fund was at risk.
• The categories were re-labeled mid-panel. The June 2024 N-MFP form revision changed the government and Treasury classification, a step this note holds outside its window. Any comparison of shares across that date mixes a real shift in holdings with a change in definitions, so the level of the government share late in the panel is not directly comparable to its level early on.