Could you please link me to "central bank" gold purchases you mentioned? Do these banks trade @CME? Sadly, I couldn't find such data that corresponds to speculators' joint impulses to sell or buy. Also, any example when they [which central banks exactly?] decreased prices when [someone, who?] wanted to sell?
The WorldGoldCouncil organisation publishes data on a term basis.
SilverInstitute does this yearly for silver.
I downloaded these since a decade, and updated into a textfile.
Here a copypaste of gold:
Central bank & other institutions
- positive figure means total net = selling
- negative figure means total net = buying
year / tonnes / average gold price that year / estimated cost per tonne based on average gold price
1997 326 $330.98
1998 363 $294.24
1999 477 $278.88
2000 479 $279.11
2001 520 $271.04
2002 547 $309.73
2003 620 $363.38
2004 479 $409.72
2005 663 $444.74
2006 365 $603.46
2007 484 $695.39
2008 235 $871.96
2009 34 $972.35
2010 -77 > -79.2(2016Q4) $1224.53
2011 -455 > -480.8(2016Q4) $1571.52
2012 -544.1 > -569.3(2016Q4) $1668.98
2013 -386.6(2014Q1) > -409.3(2014Q2) > -625.5 (2015Q1) > -623.8(2016Q4) $1411.23
2014 -477.2(2014Q4) > -588.0(2015Q1) > -590.5 (2015Q2) > -583.9 (2015Q4) $1211.71
2015 -588.4 (2015Q4) > -566.3 (2016Q1) > -576.5(2016Q4) $1160.06
2016 -383.6 > -389.8 $1250.74
2017 -371.4 > -374.8 (2019Q1) $1257.12
2018 -651.5
2010-2018 is my monitoring period, the rest I gathered together from old WGC reports linked here and there on the web.
You see that 2010-2017 have several figures, the last one is the last update = the "currently" correct one for that year.
That "2016Q4" refers to the WorldGoldCouncil report, in this case year 2016, quarter 4 report (= end of year).
You can see that some years are seriously updated in subsequent years, even 5 years later.
2013 is also a good example, in 2014Q1 they reported 386.6 tonnes bought, and the last 2016Q4 update increased it to 623.8 tonnes bought.
Just to illustrate, those that read WGC reports, and base conclusions on it, these can end up quite wrong.