Steem econometrics: net flows
Here are some results from my study of Steem value transfers. I gathered all transfers, claimed rewards, producer rewards for the week 7/20 through 7/26, and assigned them to categories of accounts based on my earlier analysis of wealth distribution.
Because 5% partitions seemed too crude, I went back and looked at 1% increments of the active user pool:
| Category | Total Wealth | Average | Minimum | Asset Distribution |
|---|---|---|---|---|
| Insider accounts (8) | total 84936467 STEEM (29.32%) | average 10617058.4 STEEM | 11.2% STEEM, 0.0% SBD, 88.7% VESTS | |
| Exchange accounts (13) | total 67679616 STEEM (23.37%) | average 5206124.3 STEEM | 72.6% STEEM, 20.4% SBD, 7.0% VESTS | |
| Inactive accounts (722333) | total 249270 STEEM (0.09%) | average 0.3 STEEM | 0.0% STEEM, 0.0% SBD, 100.0% VESTS | |
| Top 3589 accounts (1% of active users) | total 117595160 STEEM (40.60%) | average 32765.4 STEEM | minimum: 2961.157 STEEM | 14.2% STEEM, 2.5% SBD, 83.4% VESTS |
| Next 3589 accounts | total 6591614 STEEM (2.28%) | average 1836.6 STEEM | minimum: 1165.490 STEEM | 15.9% STEEM, 5.4% SBD, 78.7% VESTS |
| 2-3% (3589) | total 3174458 STEEM (1.10%) | average 884.5 STEEM | minimum: 649.643 STEEM | 16.6% STEEM, 6.6% SBD, 76.8% VESTS |
| 3-4% (3589) | total 1893272 STEEM (0.65%) | average 527.5 STEEM | minimum: 429.431 STEEM | 17.3% STEEM, 8.1% SBD, 74.6% VESTS |
| 4-5% (3589) | total 1259826 STEEM (0.43%) | average 351.0 STEEM | minimum: 288.108 STEEM | 19.3% STEEM, 11.0% SBD, 69.8% VESTS |
Using that partitioning of users, here are the net flows of STEEM (or SBD, or VESTS, converted into their STEEM values) over the week I studied:
| category | STEEM in | STEEM out | Net flow |
|---|---|---|---|
| top 1% | 3708217.9 | 3076146.6 | +632071.3 |
| next 1% | 244926.4 | 224091.9 | +20834.5 |
| 2-3% | 101981.9 | 114391.8 | -12409.9 |
| 3-4% | 86481.2 | 76042.3 | +10439.0 |
| 4-5% | 61345.7 | 57937.3 | +3408.5 |
| rest of user base | 241985.2 | 264929.4 | -22944.2 |
| exchanges | 4331801.2 | 3723788.6 | +608012.6 |
| insiders | 4.1 | 800071.2 | -800067.1 |
| reward pool | 0.0 | 391538.8 | -391538.8 |
| producer rewards | 0.0 | 47805.9 | -47805.9 |
The full data set contains all pairs of category-to-category flows, so for example I can extract where witness rewards went:
| witness category | value (STEEM) | percentage |
|---|---|---|
| top 1% | 44702.908711 | 93.5% |
| exchanges | 1914.179740 | 4.0% |
| 2 | 792.204550 | 1.7% |
| 5 | 161.822499 | 0.3% |
| 3 | 134.837677 | 0.3% |
or where content rewards went:
| category | value | percentage |
|---|---|---|
| top 1% | 187582.423665 | 47.9% |
| 2 | 58883.056978 | 15.0% |
| 3 | 30528.075941 | 7.8% |
| 4 | 22211.332076 | 5.7% |
| 5 | 18924.946431 | 4.8% |
I will be doing some more work on ways to analyze and visualize this data, but I wanted to share some early results.
Methodology notes
I extracted all on-blockchain operations of type 'transfer', 'transfer_to_vesting', 'transfer_to_savings', 'transfer_from_savings', 'claim_reward_balance', 'account_create', and 'account_create_with_delegation'. For witnesses, I had to walk their individual account history to get the 'producer_reward' virtual op.
So, the analysis excludes all activity which takes place only as virtual ops, such as internal market transactions, but (due to my choice below) those can be treated as value-neutral. It also means I count only claimed rewards, not pending rewards.
I used a fixed value for SBD and VEST conversion to STEEM, rather than trying to match each transfer with its then-current value.
Summing all net inflows and outflow resulted in an error of -0.2764 STEEM, which seems acceptably low.