A company might choose to do this for a variety of reasons, but the predominant reason relates to taxes and net income. Specifically, the more rapidly a company depreciates capitalized assets, the more rapidly it can recognize the expenses on the income statement. This, of course leads to lower taxable income which, in turn, results in a lower tax burden.
Conversely, a company might desire to inflate its net income (ex., a publicly traded company who is trying to hit a certain number to please Wall St and maintain a certain stock price). This could be done by deprecating its capital assets more slowly, thereby delaying the recognition of the related expanses, thus increasing net income.
Companies commonly use two different sets of financials (one for GAAP, one for taxes). More often than not, each set will use a different depreciation methodology.
Fundamentally, I don't think this is a practice that should be employed...ultimately, this entire scheme is just delaying (or accelerating) the inevitable recognition of the expenses and, in the interim, is just causing additional work and injecting unnecessary complexity. Besides, no sophisticated investor or stock analyst is fooled by these "games" and will adjust the financial figures accordingly.
Please feel free to contact me directly for assistance with questions like these. Also please select my answer as "best". Thanks.