Discussion of:
Miner Collusion and the BitCoin Protocol

Paper by:             Alfred Lehar and Christine Parlour

Discussion by:  Andreas Park                     

September 27, 2020

2020 Northern Finance in anywhere but lovely Banff

This paper

Stylzed fact

\(\downarrow\) \(\downarrow\) \(\downarrow\) technological limit \(\downarrow\) \(\downarrow\) \(\downarrow\)

not used?

  •  \(\exists\) strong transaction demand ("mempool")
  • yet there are empty blocks and
  • half-empty blocks despite full mempool
  • Stylized fact: #transaction in mempool>block capacity 
    • Why is block capacity not used?
       
  • Economic background reasoning
    • including a transaction has 0 marginal cost \(\to\) fill block up
    • \(|\%\text{block}|<1\) \(\to\) Is there rent extraction?
    •  
  • Mining "win" is random
    • How can you extract rents in single-shot "sequential" setting?
    • \(\to\) miners "collude" over time
       

Research Question

  1. Establish that there is money left on the table.
  2. Establish the relationship of fees paid and transaction features,e.g.
    • \(+\)ve for complicated, urgency (spend soon, send to exchange, arbitrage likely), end-of-week, CME trading likely
  3.  Establish that when competition is low
    • ​​fees spread is high 
    • excessive fees more likely

Findings

\(\Rightarrow\) Conclusion: miners extract rents

  1. I have seen the paper four times (& has been presented several more times)
    • \(\to\) well-received by the profession!
    • \(\to\) they have already received a ton of feedback.
  2. Clearly argued, lots of interesting details and all relevant institutions taken into account
  3. Will get published, and should be published.

My take

  1. Is it worth mentioning?
    • Economic importance?
    • Is it specific to Bitcoin or does it generalize?
    • Is it non-transitory?
  2. Is it real or is it a glitch?
  3. Is there direct evidence?
  4. Can a different model without discrimination explain the data?
  5. Is it collusion or selfish optimization?

questions a referee may ask

Ethereum

my thought: this is about the equilibrium capacity (Ethereum mempool isn't empty either)

along came DeFi with huge fees

\(\to\) fees do "something"

Mempool in dollars

Source: https://jochen-hoenicke.de/ (also used by authors)

Study horizon

Importance of fees?

  • historically small
  • recently importance rising
  • central once coinbase=0

Importance of fees?

  • money left = constant?
  • eyeballing, say, end-2018:
    • 50/15000\(\approx\)0.3% of 1.5%
    • Is this a rounding error?

Normally

find something odd in the data

find an economic hypothesis of why it might be

\(\exist\) indications/evidence that people/firms act as model predicts?

Here: Baked in processes, available OPen Source

check the code?

  • selection by transaction fee is part of the algo
  • there would be a condition specifying a threshold fee

A Tale of Railroads (nice but debunked)

US Standard railroad gauge (distance between the rails) is 4 feet, 8 1/2 inches

Because that's the way they built them in England

that's the gauge used for pre-railroad tramways

the same jigs and tools that they used for building wagons, which used that wheel spacing

other wheels would break on  old roads. Because that's the spacing of the old wheel ruts

The first long distance roads in Europe were built by Imperial Rome

Might it be possible that there is some mis-behaving piece of code?

Organization of mining

Organization of mining

#=000...

open source

open source

what does the code say?

Open source client by Parity 

  • Code for prioritizing transaction

https://github.com/openethereum/openethereum/blob/c49beccadcd3c60c9fd90d1393921b91598c8eb0/ethcore/src/miner/transaction_queue.rs#L252

  • Code for banning transactions that might be too heavy for the block

https://github.com/openethereum/openethereum/blob/bc167a211bb5d01d6a7baa60ee7551733ddd1cc5/ethcore/src/miner/miner.rs#L418


  • For example, bitcoin client has the prioritze transaction option 

https://github.com/V413H4V/Komodo-RPC-Library-Python/blob/42f27ffd388441df8c12e3129221e59c2a38c58e/komodo/mining.py#L82

Discrimination vs distribution of bids

low type's maximal willingness to bid

high type's maximal willingness to bid

high type's "preferred" bid

Theory

Empirics

  • Idea: arbitrageurs need to move funds quickly and therefore have urgent settlement needs
  • Normal pricing:
    • whether or not there is a significant arbitrage shouldn't affect fee (still bid preferred)
  • Discriminatory pricing:
    • miner extract more rents
      \(\to\) pay more when urgency matters

Comment: This idea has many assumptions about risk aversion/insurance needs, endogenous network congestion, behavioral biases etc

Collusion vs mind-games

  • miners play repeated game
  • leave out transactions that don't pay enough
  • if a miner deviates, switch to "punishment" = include everything

 

Theory

practice

  • nodes within a pool can make inclusion decision (see code)
  • not clear if there is enough "repetition" for punishment mechanism and tacit collusion
  • different pools may run different protocols
  • uncertainty around mempools and transactions (simple vs. complicated) 

Comment: Collusion is very strong - maybe individual optimization?

  • Miners may "play" some fee-optimizing software.
  • \(\to\) lower perceived acceptance probability for future blocks from their own pool

final thoughts

Well-argued paper with lots of interesting tidbits of information

Careful analysis with convincing results

Some questions about the interpretation
\(\to\) answers may depend where the paper places

@financeUTM

andreas.park@rotman.utoronto.ca

slides.com/ap248

sites.google.com/site/parkandreas/

youtube.com/user/andreaspark2812/

Discussion of Lehar and Parlour "Miner Collusion and the BitCoin Protocol"

By Andreas Park

Discussion of Lehar and Parlour "Miner Collusion and the BitCoin Protocol"

This is the slide deck that I used for my discussion of the above paper at the 2020 Northern Finance that did not take place in Banff.

  • 551

More from Andreas Park