In general the effective rate was much more volatile the further back you go, but now it is perfectly straight, what has changed? The fact that it is exactly three evenly spaced dips of 8-10 basis points, that stops after the March 22nd FOMC meeting is interesting.
The observed pattern is essentially (regulatory) arbitrage by banks.
1 ) One such concern is "window-dressing", in the form of temporary reductions of transaction volumes in key financial markets around reference dates resulting in the reporting and public disclosure of elevated leverage ratios. Simplified, if you prefer to report less "risky" balance sheets, you would like to show that your (overnight) liquidity is mainly with the FED. Most financial reporting is conducted at quarter end (with mid-year and year-end being the most important dates). Placing your cash with the Fed on the last day of the quarter for one night and then taking it out makes your balance sheet look less risky. That way, it looks like firms (banks) can essentially operate with less capital, boosting returns to shareholders and payouts to top executives. Lehman Brothers Repo 105 is an infamous example.
Since early 2015, which marks the beginning of the Basel 3 leverage ratio disclosure, the amplitude of swings for Swiss and Euro banks (which rely on quarter-end figures), were very pronounced, but less pronounced for UK and US banks (which use averages). Generally, it depends on the region: The US, require the ratio to be fulfilled on the basis of period averages, which adds less regulatory "pressure".
There have been a number of Basel 3 requirements and BIS reports about window dressing.
- The market risk framework: 25 years in the making; 25 April 2018
- Basel Committee consults on disclosure requirement to address leverage ratio window-dressing; 13 December 2018
- To deter window-dressing, the Fed uses a floating start date for its global market-shock. Banks are only informed of the actual date once the stress test scenarios are announced. The start date used to fall between January 1 and March 1 of the year in which the stress test was taking place. In 2018, the range was extended by the FED to begin on October 1 (it concluded the original three-month window was being gamed). The change made window-dressing more complex and costly. Needless to say, there have always been attempts to "game" stress tests.
2 ) Another contributing factor is arbitrage available to foreign banks in US money markets due to not having to pay deposit insurance. See for example this article. However, I think since the pattern vanished, it was mainly the first point as opposed to the second point that really explains it.