[tptacek]

It's not just that losing 92% of their case doesn't cost the prosecution at sentencing. It's that those charges continue to pay dividends to them, even though they're counterfeited by the actual trial.

It seems like one good inoculation against overcharging would be for damages and facts establishing conduct to be fenced off by which charges succeed and which fail, so that failing to convict on a charge puts part of the sentence at risk, as opposed to filling up some freakish "community chest" of sentencing accelerators.

[mpyne]

Even the Granick analysis you quoted earlier came to "over a year" for the probably guideline calculation, not 35 or 50. And it never talked about additional charges adding to the sentence in the "sentencing" portion.

She did talk about each additional conviction adding to the maximum possible sentence for a plea bargain, but then said that the guideline calculation is still the same.

She linked a PDF of hers, and I tried to read it, but ironically it requires HTTP authentication to some journal archive which I haven't paid for and don't have access to.

I've also tried going through the 2011 Federal sentencing guidelines to confirm what you gleaned from Granick, but to be honest they're too byzantine for me to get through in a few minutes with no training.

I guess what I'm trying to say is that I'm not sure how you got from Granick to what you're talking about here. Perhaps I missed a linked article from the site hosting her original article?

Either way I don't see how it would make sense to credit '# of charges' alone in sentencing as it is well-known that in general the prosecution has to bring all charges it can prove related to the case, if only because of double-jeopardy.

[pseingatl]

I haven't seen the Granick analysis. But it's easy to do the guidelines calculation. The base level for fraud is 6, and there is an 18 level enhancement when the amount in question is between 2 and 7 million dollars, as the government claimed. That yields, with no criminal history, a guideline level of 24 and a sentence of 57-71 months if there were no other enhancements. There almost always are. You can drop 2 levels if the defendant "accepts responsibility" which may or may not have happened. There would be 2-3 years of supervised release (no computer) and a fine within statutory limits. The danger with multiple counts is that the sentencing judge may decide to impose consecutive punishments which would exceed the guideline range. In this particular case, which, make no doubt about it, was a political case, Ortiz' office might seek an upward departure because of the defendant's lack of remorse, the PACER incident, and a perceived "danger to the community."

[tptacek]

"related conduct, even aquitted conduct, is part of the sentencing calculation."

[mpyne]

But why wouldn't that be part of the sentencing calculation?

Let's say I get charged under, say, 100(c), with elements of proof [A, B, C] and 100(d) with elements of proof [D, E, F]. They convict me on all of A, B, C (and therefore find me guilty of violating 100(c)). They find I actually broke D and E too, but not F, therefore I'm not guilty on 100(d).

If D and E are related to 100(c) why should a judge not take that into consideration? The jury confirmed those elements of the charge were true beyond a reasonable doubt.

I have heard from somewhere that judges may consider elements that were not confirmed by the jury as long as there's still a preponderance of evidence to support. That could be an issue I suppose, if you suspect the judiciary is on average corrupt. But even that still passes the "common sense" test IMHO.

Either way, what I got from Granick was that the worst element of the sentencing calculation by far was the loss calculations (which could essentially easily be "blank check"ed by MIT), not that the judge is allowed to consider related conduct proven by the defendant.