There's a few different ways of defining what you mean by 'Calories' in food.  There's the obvious physical/chemical definition of energy stored in the food, but I had an objection when I explained that [warmer food naturally has more Calories than cold food.][1]  There's also the measure of absorbtion, or ['digestibility', which can be determined by what nutrients pass through the body unabsorbed][2].

The [US FDA has guidelines][3] that basically suggest for their labeling requirements, it's based on estimates of calories in the component parts of the food:

> L12. Is it necessary to include a calorie conversion footnote which states that fat, carbohydrate, and protein furnish 9, 4, and 4 calories per gram, respectively?
>
> Answer: No, the use of that footnote is optional. 21 CFR 101.9(d)(10)

But they also have rules on rounding:

> L31. What are insignificant amounts of nutrients?
>
> Answer: These are the amounts that are permitted to be shown as zero on the Nutrition Facts label (e.g., less than 5 calories may be expressed as 0 calories) except that for total carbohydrate, dietary fiber, and protein, it is the amount that can be declared as “less than 1 g” on the Nutrition Facts label. 21 CFR 101.9(c)

It seems that less than 0.5g is required to be listed as '0g', even though they call for rounding at other times:

> L45. How should trans fatty acids be listed?
>
> Answer: Trans fatty acids should be listed as “Trans fat” or “Trans” on a separate line under the listing of saturated fat in the Nutrition Facts label (see figure). The word “trans” may be italicized to indicate its Latin origin. Trans fat content must be expressed as grams per serving to the nearest 0.5-gram increment below 5 grams and to the nearest gram above 5 grams. If a serving contains less than 0.5 gram, the content, when declared, must be expressed as “0 g.” (21 CFR 101.9(c)(2)(ii)).

...

Some items are allowed to take into account absorbtion (eg, Olestra, the fat that humans can't absorb), but there's also personal differences -- not all people absorb nutrients the same way, due to food intollerances.  There's also the question of how [cooking affects absorbtion][4], as some [studies have shown different digestibility between cooked and uncooked items.][5]  Of course, nutrition labels don't have a way to explain that there's different effective Calories depending on how something's prepared.


Personally, I go with the assumption that they're inaccurate estimates.  It's my understanding that the FDA doesn't even care unless the values are more than 20% off.  As such, there's no incentive for companies to test their product -- if they found out their values were wrong, they might be forced to correct them, and doing so could put them at a disadvantage to other groups that were 19.5% wrong in their values.  Also, as most food products are agricultural, there's going to be differences between tomatoes from one plant to another, and they just average these things out ... but we have no idea what the variance is for any of the values.

  [1]: http://cooking.stackexchange.com/questions/66/how-to-calculate-the-calorie-content-of-cooked-food/500#500
  [2]: http://www.fspublishers.org/ijab/past-issues/IJABVOL_5_NO_1/26.pdf
  [3]: http://www.fda.gov/Food/GuidanceComplianceRegulatoryInformation/GuidanceDocuments/FoodLabelingNutrition/FoodLabelingGuide/ucm064904.htm
  [4]: http://cooking.stackexchange.com/questions/66/how-to-calculate-the-calorie-content-of-cooked-food
  [5]: http://www.beyondveg.com/tu-j-l/raw-cooked/raw-cooked-2a.shtml