This document summarizes a study of the relationship between bid-ask spreads in the retail foreign currency exchange market in Moscow and spreads in the interbank FOREX market. The study found:
1) Retail spreads significantly correlate with and are indirectly impacted by volatility and trading results in the interbank MICEX market.
2) Retail spreads are more elastic during periods of low volatility and less elastic during high volatility.
3) There is intraday but not interday hysteresis in retail spreads. Spreads are highest in the morning and decline through the afternoon.
4) Estimates support the Kyle-Obizhaeva two-factor model, with positive correlations between retail spreads and interbank market
Microteaching on terms used in filtration .Pharmaceutical Engineering
Bid ask spreads in the retail currency exchange offices (Master's Thesis)
1. Bid-Ask Spreads In The Retail
Currency Exchange Offices
Individual Project
And
Their Relationship
With Interbank Forex
Market
2. Introduction
• The goal of the project is to analyze the
USD/RUB bid-ask spreads on the retail exchange
market in Moscow and its relationship with
MICEX results
• I found that retail spreads significantly correlate
with MICEX trading results. However, this
correlation is complex and indirect.
• I found some evidence that two-factor model
proposed by Kyle-Obizhaeva study (2014) are
relevant to the retail exchange market and
displays some prediction power
3. Introduction
• As the most of retail factors are usually unobservable,
using common market data could be valuable for studies.
• Even assuming additional retail transaction costs
(Yuanchen Chang 1996) I believe, the two-factor model
could predict retail market spreads to a considerable
degree.
4. Data
• I collected a panel data of around 800 Moscow retail currency exchange offices in period
of January,14 2015 to May,15 2015 by on-line quotation resource quote.rbc.ru.
• I built panel data of 10-minute snapshots of a citywide retail currency exchange market.
• For each snapshot I calculated geometric mean for the smallest 10 bids (2,5% percentile)
as a sample for of current market risk (Bollerslev, 1994)
5. Methods
I used different frameworks to explore each of
the following:
• correlation among MICEX trade size, volatility
and retail spreads
• possible seasonality in retail spreads
• hysteresis of retail spreads
6. Methods
I examined four different functional forms: linear, squared, square rooted and
logarithmic:
ln 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑒𝑡𝑎𝑖𝑙 𝑠𝑝𝑟𝑒𝑎𝑑 = 𝛼 + 𝛽 ∙ ln 𝑀𝐼𝐶𝐸𝑋 𝑟𝑒𝑡𝑢𝑟𝑛 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦
I tested seasonality in three different ways:
1. Estimation of the distribution of extreme values
2. the exploration of possible cyclic component in time-series
3. I assessed the distribution of standard deviations
I estimated both equations forms to test the Kyle-Obizhaeva findings on the
dependence among volume, volatility and spreads on the market.
First, “the short-form”:
𝑙𝑛
𝑅𝑒𝑡𝑎𝑖𝑙 𝑠𝑝𝑟𝑒𝑎𝑑 𝑏𝑎𝑠𝑒 𝑝𝑜𝑖𝑛𝑡𝑠
𝜎
= 𝛼 + 𝛽1 ∙ 𝑙𝑛(𝑇𝑟𝑎𝑑𝑒 𝑣𝑜𝑙𝑢𝑚𝑒 ∙ 𝜎) (1)
The second, “the long form”:
𝑙𝑛 𝑅𝑒𝑡𝑎𝑖𝑙 𝑠𝑝𝑟𝑒𝑎𝑑 𝑏𝑎𝑠𝑒 𝑝𝑜𝑖𝑛𝑡𝑠 = 𝛼 + 𝛽1 ∙ 𝑙𝑛(𝜎) + 𝛽2 ∙ 𝑙𝑛(𝑇𝑟𝑎𝑑𝑒 𝑣𝑜𝑙𝑢𝑚𝑒) (2)
Where,
Retail spread bp — retail as a base points of weighted average exchange rate
Ϭ — sigma — standard deviation of the rate for the day/last hour
Trade volume — (rate · USD trading volume) or volume in rubles of trade for the day/10-minutes
7. Results 1/3
The logarithmic form is the best estimation for
the retail spread elasticity over MICEX’s
volatility
10. Conclusions
• The retail market is more elastic for low volatility
while higher levels of price deviations made spreads
less elastic
• There is no hysteresis of retail spreads between
trading days, while intraday statistics shows 30-40
minute trade volume lagging in spreads
• The retail dealers tend to perceive morning operations
with higher risk of potential losses and keep their
spread wide until 11:00. Just after 11:00, spreads
abruptly fall and gradually declining to the day’s
minimums about 13:00-15:30.
11. Conclusions
• My study supports the Kyle-Obizhaeva findings and show significant
evidence of strong correlation between retail spreads and interbank
trade volumes and volatility. I found that such interconnection
persists even between different markets of the same asset.
• I revealed positive correlation of retail spreads with volatility on
MICEX.
• My estimations disclosed positive correlation within average retail
spread and MICEX trading volume:
– increase in trade volume often leads to significant shift in exchange
rate. Giving the positive correlation between volumes and MICEX
returns. Retail dealers apprehend higher risk of exchange rate shift
and boost their spreads, because of one-hour rate hold-up period.
– Another explanation: trade volume on interbank currency exchange
mirrored in retail trade volumes on a limited basis due to cash/non-
cash transactions costs. Thus, retail exchange offices may not actively
hedge their currency positions on MICEX
12. Modeling
• I tried to apply revealed estimation coefficients of
Kyle-Obizhaeva equation to predict spreads in
the last week:
Corr.
coefficient:
~0,36
R2=0,32