That's a very good question. Again, these are frequencies. But let me talk about the actual number of accidents and the difference between FRA reporting....
I would agree with Mr. Miller's comments in terms of recommendations. We can go further on the panel's recommendations on the data capturing and how we use the data. Frankly, the FRA data we are required to report in the United States gives us much greater granularity around accidents and the cause of accidents.
For clarification, an FRA accident is an accident that roughly costs more than $8,500 to rectify, not including lighting damage. That's just the cost of repair. Anything less than that is a non-FRA train accident.
The chart on FRA train accidents in 2007 that you see represents 90 accidents on the Canadian Pacific Railway that cost more than $8,500 to rectify. Of those 90 accidents, about 45 were in yards and 45 were on line of road. This is the tip of the iceberg. When you go below the iceberg to be more consistent with the TSB reporting, there are approximately 1,100 non-FRA accidents—the accidents that cost less than $8,500 to rectify.
To your question of whether we think we have levelled off, when you look at the total accident community, of the 1,200 accidents, including the FRAs, I believe you'll see a trend of reduction there. And I believe we can bring that level of accidents down further. As you see, in 2007, according to this matrix, we essentially had two accidents per million train miles.
Our objective within four years is to get to one accident per million train miles, so essentially to see another 100% improvement in that level. Is that going to be difficult? I believe it is. It's an aggressive target. It is one that is discussed and set with our board of directors, and they're encouraging us. So the next question is how you do it. I do believe that the panel's recommendations, some of the technologies I shared with you, and the human factor of what we're talking about are the key drivers in how we're going to drive those accidents to a much lower level than we see today.