A study of 41 billion trade reports since 2006 reveals a clear winner from retail price improvement. Hint: it's not the retail trader.Salami Slicing from Wikipedia:
We counted all eligible trade reports from stocks listed on NYSE, ARCA, AMEX, and Nasdaq from January 3, 2006 through July 25, 2012. To be eligible, the trade price had to be greater than $2. To qualify as a subpenny trade in this study, the price had to require more than 2 decimal places, and not be priced exactly at 1/2 cents. See this report for more information on subpenny trades, who benefits, who loses, and what the SEC has stated and ruled on this important topic. We also published a paper on subpenny trades in Apple stock. 
The pie chart below shows the distribution of the subpenny trades in our study. A full 28% (red slice) of these trades had a price improvement at the lowest permissible amount: 1/100th of a cent ($0.0001) per share.
The other slices show that 26% (green) of subpenny trades had a price improvement benefit between 2/100 and 10/100ths of a cent, and 46% (blue) had a benefit between 11/100 and 49/100 of a cent. 
Note that over half of all price improved trades
(54%) involve an economically insignificant amount (1/10th of a cent).
In the SEC's own words:
The next chart further details the distribution of subpenny trades. The two spikes in the center represent prices with fractions of a cent of 99/100 and 1/100  both of which are 1/100 of a cent away from the nearest whole cent. The sum of these two spikes make up the red slice in the pie chart above. Note how the data in the left half of the chart is a near mirror image of the data on the right half of the chart. That is because internalizers (wholesalers, those who buy orderflow from retail brokers) give buyers the same price improvement they give sellers. Also note that the left scale is in billions of shares, and the right scale is in trillions (yes, trillions) of dollars. 


The table below shows the combined values of the two sides of the mirror image above. In other words, we sum the number of shares and transaction cost (value) for trades with subpenny prices of 1/100 and 99/100 and place the results in row 1. Row 2 is the combination of subpenny prices of 2/100 and 98/100 and so on. The last 3 columns show the estimated benefit and harm to the 3 parties impacted by subpenny pricing. These 3 parties are discussed in detail here, but for easy reference:
The first entry (row 1) tells us that 204 billion shares worth $5.1 trillion executed just $0.0001 away from the nearest cent. The group labeled Investor A benefited in these transactions by a total amount of $20.4 million. This is called retail price improvement. The group labeled Investor B lost at least $2 billion from these transactions. Finally, the group labeled INT (internalizers) received a benefit of at least $2 billion.


You can download a pdf of this data here. 

The Final Score:
The next chart shows the growth of subpenny trading between January 2006 through July 25, 2012. The line is a plot of subpenny trades as a percentage of all trades. We used a monthly average of daily percentages to smooth out fluctuations. 



The next two charts show the impact of the Knight event that
occurred on August 1, 2012. Knight's absence caused a significant reduction in trades priced in fractions of a cent,
which returned after Knight recovered. 

1. Chart showing % of share volume of trades with prices ending in 1/100 or 99/100ths of a penny (e.g. 50.4599 or 81.2301)  2. Chart showing % of share volume of trades with prices ending in a subpenny fraction (excluding 1/2 cent because those can come from other sources). 