Common sense says that if you leave a more valuable item out, it is ceteris paribus more likely to be stolen. If we're asked why, I think we'd say: "The temptation is bigger." But this can't be the whole story. For while the reason for stealing is proportional to the item's value vt to the potential thief, the moral reason not to steal seems proportional to the value vo to the owner (after all, surely it's about equally bad to steal two items that are worth 50 thalers each than to steal one that's worth a hundred). Moreover, roughly, vt is proportional to vo (with a convex tapering off at the high end of vo as very valuable things are hard to fence). So we might expect that as we double vo, we roughly double the strength of the potential thief's reason to steal and roughly double the strength of the moral reason not to steal, and we might expect the two effects to cancel out, leaving the chance of theft unchanged. Compare theft to bribery. One has no greater a moral reason to refuse a large bribe than to refuse a small one, at least if the briber can afford it. In the bribery case, there is nothing surprising about people being more likely to take greater bribes.
The above analysis, however, neglects the fact that one also has a non-moral reason not to steal from the chance of getting caught and punished. And the punishment is not directly proportional in intensity to vo: one doesn't go to jail for ten thousand times as long for stealing ten million dollars as for stealing a thousand dollars, I assume. Moreover, there is an initial non-proportional disvalue just in getting caught, dealing with the legal system, etc.
However, I do not think the lack of exact proportionality in punishment fully accounts for our intuition here. For, I think, we may also have the intuition that more valuable items are more likely to be stolen in the case of thefts where the thief thinks he can't get caught. We just see the more valuable items as "more corrupting". (Actually, maybe, the function is somewhat more complex. Items of very low value may get stolen more easily "because it doesn't do any harm", and items of very high value may get stolen more easily "because they corrupt", while items of middle value might be less likely to be stolen. I will ignore the extreme low end of value.)
The "more corrupting" intuition can, I think, be accounted for in this way. We recognize an intrinsic disvalue in acting viciously or becoming a vicious person. But the disvalue is more binary: being a thief versus being a non-thief; committing this theft versus not committing this theft; being a criminal versus being innocent. Catholics might make a ternary distinction: acting rightly versus sinning venially versus sinning mortally. But still there is a lack of proportionality.
It may be quite beneficial to have such a lack of proportionality—it keeps us from first stepping over the threshold of a new kind of moral evil, such as theft. If we were willing to commit small thefts on the grounds that they are not really morally all that wicked, we would be likely to slide towards greater ones. So it makes sense for there to be a bar against embarking on a new kind of wickedness. And then the greater temptation helps overcome that initial bar. If so, then the felt disvalue of committing a theft when one has not done so as yet is: B+cvo, where B is the initial bar against "becoming a thief" and c is some proportionality constant. If that is the right story, then we might predict that people who see themselves as thieves are as likely to steal small amounts as great when they do not expect to get caught, while people who do not see themselves as thieves are more likely to fall for a greater amount.