One component of the upcoming hard fork 21 is the Economic Improvement Proposal (EIP), explained in technical detail here on the @steemitblog and here in layman's terms by @justineh.

The goal of these economic changes would be to move Steem closer to delivering on the promises stated in the whitepaper of unearthing high quality content by making it more profitable for people to actively curate, and less profitable to self-vote and delegate one’s Steem Power to bid bots.

(As stated by Steemit, Inc. employees on the @steemitblog)

But how will we know if the EIP is a success? What is the current situation RE the problem behaviors identified by the authors of the EIP?

For the last month I have been procrastinating finishing an analysis post in which I hoped to provide benchmarks to judge the success of the EIP. In all honesty, it has been giving me anxiety and every time I start to dive back into the numbers it has depressed me. I am scrapping the larger intended post, but I still want to share one aspect of research I already completed RE vote selling.

I do not intend this post to be an argument for or against the EIP, or even to be an argument for or against vote selling, but rather to provide what I believe to be crucial background information that has so far been missing from the @steemitblog updates and the various HF21 related posts recently featured on steemit.com.

NOTE RE "BID BOT" TERMINOLOGY

For the rest of this post, the term "bid bot" will be used as short hand for a variety of different forms of vote selling/buying, many of which are not technically "bid bots". See Methodology below for a more detailed explanation of what is included and why, but in short I have included bid bots, paid promotion services and delegate-for-upvote bots all under the short hand label "bid bots".

THE "BID BOT" TERM DOES NOT INCLUDE DIRECT VOTE SELLING.

The MinnowBooster and Smart Market direct vote selling marketplaces are handled separately in the following analysis.

Benchmark #1 - % of SP controlled by bid bots and direct vote sellers

Active SP is the net SP controlled by all accounts that voted at least once in the 30 days prior to July 15th (owned SP + delegated SP - outgoing delegations).

CHANGE OVER TIME @crokkon conducted a similar analysis almost exactly one year ago that found 21% of active (30 day) SP controlled by bid bots, so we have seen a nearly 25% increase since then.

*Approximate Vote Sellers SP

Total reported vote worth available on the direct vote markets converted to ~ voting SP (MinnowBooster reported $56 total vote worth of all vote suppliers; SmartSteem/SmartMarket reported $172 vote strength in vote market; voting SP calculated at the then current value of $.24 Steem using the SteemNow.com upvote calculator and rounded down .

This is likely not a perfect method for calculating the total SP held by vote sellers, because I believe the $ value reported by the vote markets depends on the current vote strength of the voters. I have performed this calculation several times over the past month since I made the above chart and have not come up with a value less than 16 million SP. At times the number was north of 16.5 million SP, so I believe 16 million SP is a conservative baseline estimate for combined MinnowBooster (~4 million SP) and SmartMarket (~12 million SP) vote markets.

If anyone associated with @minnowbooster or @smartsteem (hi @therealwolf & @buildteam) can provide a more informed number of total SP held by users who have made their vote available on the vote markets it would be appreciated :)

Also please tell me in the comments if I am missing any vote selling markets.

For the rest of this analysis I am only focusing on the bid bot SP, because the method above used to approximate SP available for direct vote buy cannot be used to segment curation or author reward on historical data.

Overview of Author, Curator and Beneficiary Reward

Benchmark #2 - % of curation reward controlled by bid bots

I chose to look at % of curation reward controlled by bid bots rather than r-shares awarded, both because of the proposed change to 50/50 curation under the EIP and also because r-shares do not capture reward that may be lost from front-running votes. It is interesting to note that the bid bot curation share is slightly higher than the share of SP controlled by bid bots (27% vs. 26%), even accounting for front-running votes removing some of that curation share. Bid-bots use available SP resources more efficiently than human actors, and a small loss in curation share from front-running votes is evidently not enough to offset that advantage.

Although I will not be completing my research into self-voting and flag behavior, I think vote selling is the best metric to judge the success of the EIP by in any case.

Self-voting behavior in a system that allows essentially unlimited anonymous accounts and anonymous transfer of stake cannot be truly measured, but @crokkon recently published a detailed analysis on self-votes and low diversity voting which gives us as good of a benchmark to judge self-voting by as anything I could come up with.
I will summarize @crokkon's conclusions here:

• 4% of all votes by value are currently direct self votes.
• ~ 7-8% of all votes by value are currently coming from voters that spend 80% or more of their vote value to 3 or fewer authors
• ~10-12% of all votes by value are currently coming from voters that spend 80% or more of their vote value to 5 or fewer authors
  (credit to @crokkon for these conclusions and see @crokkon's self-voting analysis post for methodology)

I would caution against using self-voting benchmarks to judge the success of the EIP however, because it is possible any change in the % of obvious self-voting may just reflect spreading stake out into more accounts to obscure the behavior. As we saw above, vote selling commands a much larger share of the reward pool than obvious self-voting in any case, and will provide a better benchmark for that reason as well as the difficulty in assessing true change in self-voting behavior.

Flags will be the reason self-voting or vote selling decreases, if either of those things happens under the EIP. It might therefor be instructive to look at flag statistics pre and post HF21 to understand why the EIP is succeeding or failing, but not to determine if the EIP is succeeding or failing.

Methodology

I used the MySQL database and SQL query to pull the data, with some metrics derived in Excel. If you have any questions about the exact queries and formulas used, ask me in the comments and I will share the specifics.

Steem per vest number used to calculate SP from vests: 0.0005032 (this was the number when I was conducting the analysis back in July).

• I chose to convert vests to SP to make the results more easily human readable and understandable. The actual number of Steem per vest is constantly increasing at a very small rate. See steemd.com "S/V" value for the current rate. If you are interested in the actual number of vests and not a very close approximation of SP, simply divide the SP figures given above by 0.0005032.

Bid Bot List

• I started with the list of bid bots and paid post promotion services at https://steembottracker.com/ and added the list of paid vote bots put together by @crokkon.
• I filtered out all accounts that had not cast any votes in the analysis period (June 1st - July 15th).
• Then I went through the list one by one and looked at the account activity to verify it was some form of pay for upvote account. I removed all accounts that I could not definitively verify were engaged in some form of payment for upvote. The largest account by SP that I removed from @crokkon's list was @busy.pay (524,995 SP).

Final Paid Vote List

'aafrin', 'a-bot', 'alfanso', 'alliedforces', 'appreciator', 'authors.league', 'automation', 'barbro', 'bdvoter', 'big-whale', 'blockgators', 'boomerang', 'booster', 'botcoin', 'brandonfrye', 'brotherhood', 'brupvoter', 'buildawhale', 'bumper', 'byresteem', 'chronoboost', 'coolbot', 'drotto', 'earthnation-bot', 'ebargains', 'echowhale', 'edensgarden', 'emperorofnaps', 'estream.studios', 'flagship', 'followforupvotes', 'fresteem', 'friends-bot', 'giftbox', 'honestbot', 'hottopic', 'hugewhale', 'hugo4u', 'inciter', 'isotonic', 'jerrybanfield', 'joeparys', 'lays', 'lost-ninja', 'lovejuice', 'lrd', 'luckyvotes', 'make-a-whale', 'mercurybot', 'microbot', 'minnowbooster', 'minnowhelper', 'minnowpond', 'minnowpowerup', 'minnowvotes', 'mitsuko', 'moonbot', 'morwhale', 'msp-bidbot', 'msp-creativebot', 'msp-lovebot', 'msp-shanehug', 'ocdb', 'oceansbot', 'oceanwhale', 'onlyprofitbot', 'originalworks', 'peoplesbot', 'postdoctor', 'postpromoter', 'profitbot', 'profitvote', 'promobot', 'pwrup', 'qustodian', 'redlambo', 'red-rose', 'resteemable', 'resteembot', 'resteemr', 'rewards-pool', 'rocky1', 'schoolofminnows', 'sct.voter', 'shares', 'sharkbank', 'siditech', 'singing.beauty', 'sleeplesswhale', 'smartmarket', 'smartsteem', 'sneaky-ninja', 'spydo', 'steemdiffuser', 'steemlike', 'steemmonsters', 'steemthat', 'steemvote', 'steemyoda', 'stef', 'sunrawhale', 'superbot', 'sureshot', 'swiftcash', 't50', 'th3voter', 'thebot', 'therising', 'thundercurator', 'tipu', 'tisko', 'tommyhansen', 'treeplanter', 'triplea.bot', 'upme', 'upmewhale', 'upmyvote', 'upvotewhale', 'upyourpost', 'whalecreator', 'whalepromobot', 'withsmn', 'zapzap'