Monday, August 4, 2008

Predicting End-of-Day SPY Volume

As of 12:40 p.m. EST today, SPY volume had totaled approximately 104.6 million shares. I read an article on the web this morning declaring this to be a "low-volume day." As shown below, volume is heaviest in the mornings and afternoons, and is fairly erratic bar-to-bar. This pattern is very typical, complicating end-of-day estimation. So how does one reasonably assess the low-volume claim?


Below, I have totaled the volume data depicted above (nine days -- in my TS Windows I use 20-days), and then smoothed the result using a polynomial regression to form a "typical volume day." By dividing actual volume at any point by the typical cumulative percentage of total volume matching that period, one can thus complete the end-of-day estimation. The closer to the actual close, of course, the more accurate the estimate will be.


For example, if today holds to the typical pattern, the market should have put in about 56.8% of total volume through 12:40 p.m. Therefore, total end-of-day volume is estimated as 104.6M / 56.8%, or 184.3M shares. In comparison, five-day average total SPY volume has been 269.5M shares, and twenty-day average volume has been 324.5M shares. Looks like a relatively low-volume day indeed!

Although it's beyond the scope of this article, sophisticated readers can take this analysis further by using standard deviation z-scores to more quantitatively define what constitutes "low" and "high" for a given market.

Closing Note: It looks like actual end-of-day SPY volume came in at about 184.1M shares during regular session hours, and 188.2M shares including the after-market.



I have included the cumulative total daily volume divisors in five-minute bars below for your own use and future reference. May be a good one to bookmark.

5-Min Bar / Time (EST) / Cum Vol %

01 / 09.30 / 002.9%
02 / 09.35 / 005.6%
03 / 09.40 / 008.1%
04 / 09.45 / 010.4%
05 / 09.50 / 012.7%
06 / 09.55 / 014.8%
07 / 10.00 / 016.9%
08 / 10.05 / 018.9%
09 / 10.10 / 020.8%
10 / 10.15 / 022.6%
11 / 10.20 / 024.3%
12 / 10.25 / 026.0%
13 / 10.30 / 027.6%
14 / 10.35 / 029.2%
15 / 10.40 / 030.7%
16 / 10.45 / 032.1%
17 / 10.50 / 033.5%
18 / 10.55 / 034.9%
19 / 11.00 / 036.2%
20 / 11.05 / 037.4%
21 / 11.10 / 038.7%
22 / 11.15 / 039.9%
23 / 11.20 / 041.0%
24 / 11.25 / 042.2%
25 / 11.30 / 043.3%
26 / 11.35 / 044.4%
27 / 11.40 / 045.4%
28 / 11.45 / 046.5%
29 / 11.50 / 047.5%
30 / 11.55 / 048.5%
31 / 12.00 / 049.5%
32 / 12.05 / 050.4%
33 / 12.10 / 051.4%
34 / 12.15 / 052.3%
35 / 12.20 / 053.2%
36 / 12.25 / 054.1%
37 / 12.30 / 055.0%
38 / 12.35 / 055.9%
39 / 12.40 / 056.8%
40 / 12.45 / 057.6%
41 / 12.50 / 058.5%
42 / 12.55 / 059.3%
43 / 13.00 / 060.2%
44 / 13.05 / 061.0%
45 / 13.10 / 061.8%
46 / 13.15 / 062.6%
47 / 13.20 / 063.4%
48 / 13.25 / 064.2%
49 / 13.30 / 065.1%
50 / 13.35 / 065.9%
51 / 13.40 / 066.7%
52 / 13.45 / 067.5%
53 / 13.50 / 068.3%
54 / 13.55 / 069.1%
55 / 14.00 / 070.0%
56 / 14.05 / 070.8%
57 / 14.10 / 071.6%
58 / 14.15 / 072.5%
59 / 14.20 / 073.4%
60 / 14.25 / 074.3%
61 / 14.30 / 075.2%
62 / 14.35 / 076.1%
63 / 14.40 / 077.1%
64 / 14.45 / 078.1%
65 / 14.50 / 079.1%
66 / 14.55 / 080.2%
67 / 15.00 / 081.3%
68 / 15.05 / 082.5%
69 / 15.10 / 083.7%
70 / 15.15 / 084.9%
71 / 15.20 / 086.3%
72 / 15.25 / 087.7%
73 / 15.30 / 089.1%
74 / 15.35 / 090.7%
75 / 15.40 / 092.3%
76 / 15.45 / 094.0%
77 / 15.50 / 095.8%
78 / 15.55 / 097.8%
79 / 16.00 / 100.0%

2 comments:

Joshua said...

How do you have 79 5-minute bars? It should be 78. ( 390 minutes / 5 = 78 ) . Assuming of course you are using the NYSE session time.

Jeff Pietsch, CFA said...

Hi Joshua,

Good question! This was a ten-day sample generated from the MSN Finance site. I'm away from my research station today. In the meantime, I'll speculate that the last "bar" is empty in time, but captures on close orders.