What Was the Fed Doing in late-2007/early-2008?

This morning I stumbled across an article published by the St. Louis Fed back in 2013, discussing whether the Fed is monetizing the Government Debt. The point itself is quite weak, essentially boiling down to "Bernanke says no".

However, I found the graph provided somewhat peculiar:

Was the Fed really holding up to a fifth of the outstanding securities (as shown above), especially when concerns about China being the biggest owner of U.S. debt were still ripe? But more importantly, why did the proportion held by the Fed fall so rapidly in 2008 only to be increased again by later QE programs?

The article refers to research done by St. Louis Fed vice president David Andolfatto with a colleague, so I tweeted Dr. Andolfatto to ask. His response:

@joakimbook @stlouisfed @BobMurphyEcon Look at the denominator.

— David Andolfatto (@dandolfa) February 9, 2016

So I looked. And I am even more surprised over what I found.

Normally, Fed Monetary Policy is conducted by the FOMC specifying a Fed Funds rate, which is obtained by selling and buying securities in Open Market Operations; when the Fed buys securities from the private sector, it injects money into the economy, and puts downward pressure on the Fed Funds rate; when the Fed sells securities to the private sector, it takes money out of the economy, putting upward pressure on the Fed Funds rate. Easy enough.

As you can see from this graph, Fed gradually cut rates beginning in mid-2007 as a response to the slowdown in economic indicators, more aggressively in early 2008. In fact, during the first quarter of 2008, the Fed cut rates by more than two percentage points (200 bps). Naturally, I expected to see the Fed rapidly increasing its holding of Treasury bonds and buying aggressively to push the Fed Funds rate down.

But that's not what the data says.

Data from FRED Database, St. Louis Fed
Fed Holdings = TREAST

US Debt = GFDEBTN

In fact, from its top of 2007:Q2 to its bottom in 2009:Q1, the Fed holding of Treasury was reduced by some 40% in absolute numbers. And during the heavy rate-cutting in early 2008, the largest quarter-by-quarter reduction of Fed's holding of Treasury (-22%) occurs. Precisely when the most purchases of securities should happen, the Fed reduced its holdings the most.

Were the Fed not aware of what was happening? Former chairman Ben Bernanke, in his autobiography The Courage to Act says otherwise, commenting on one of FOMCs decisions to cut rates in the fall of 2007, months before the largest reduction:

The downside risks are quite significant, if the housing situation, including prices, really deteriorates (p. 170).

What about Andolfatto's tweet, that the increase in available Treasury Bills can explain it? Not really. The really big quarter-by-quarter increase in public debt (and Treasury Bills available) did not occur until 2008:Q3 where the average increase was $400-500bn/quarter. Before that, the stable, almost-linear increase amounted to some $100bn/quarter for a very long time. And since the actual fall in net Fed holdings occurs irrespective of how much debt the Treasury issues, the question is still on the table.

What if it's just the market value of Treasury bills falling, perhaps because of massive sell-off by investors? That would indeed explain the rapid drop in absolute-value Treasury Bills (presumably they are accounted for in dollar terms?) held by the Fed - but it cannot explain the proportion held by the Fed (from around 10% in my chart and 20% in Andolfatto to 4% in mine and 8% in his), because they're the same unit; if the nominator and denominator falls by the same percentage point, the ratio remains the same.

Besides, those numbers don't add upp with Andolfatto & Li's initial graph above. From where do they get their stable "18-20% of outstanding Treasuries" since essentially 2000?

I suspect that I have made some major mistake here, that explains all this, and I'd be happy for any input.