Sum-of-the-parts valuation of USX Corporation against its 1990 market capital, using steel and energy peer multiples.
USX was two companies inside one holding structure: U.S. Steel, domestic, cyclical and shrinking under import competition, and Marathon, an international energy business that was growing. The market priced the whole. The question was what the parts were worth.
The method is Berger and Ofek. Value each segment at the median multiple of its own pure-play peers, add the segments, and compare that imputed stand-alone value against the capital the market actually assigned to the parent.
Steel and energy traded on different multiples for good reason. The steel comparables carried negative earnings metrics through 1990, which forced the valuation onto sales and book measures rather than earnings. Energy peers supported earnings multiples.
Segment-level 1990 multiples applied to USX revenue and asset bases gave an imputed total capital of $15,655m. Actual total capital was $13,997m.
The gap is $1,658m, or 10.6% of imputed value. In log terms the discount is -0.111. Running the calculation across the 1990 price range rather than a single point puts the discount between roughly 4% and 17%, with the midpoint of the range landing at 10 to 11%.
That midpoint sits inside the 10 to 15% band Berger and Ofek find for diversified US firms, which is the useful part: the result is not an outlier that needs explaining away, it is the ordinary cost of the holding-company structure.
A 10% discount is not automatically an argument for separation. It is an argument for asking whether the diversification is buying anything. For USX the case for keeping both was capital allocation across uncorrelated cycles. The case against was that the market could do that allocation itself, cheaply, by holding both stocks.
The honest limit of this analysis is the peer set. Steel comparables in 1990 were distressed enough that the multiples carry wide error bars, which is why the range matters more than the point estimate.