Some commentators argue that US stocks are more expensive than suggested by market price/earnings (P/E) measures because corporate margins are unusually high and should revert to average over coming years, depressing earnings.

The first chart shows gross and net corporate margins, calculated from national accounts data. The gross measure expresses the sum of profits, interest and depreciation as a percentage of corporate gross domestic product. Gross margins are at a record high and would have to fall by 19% to return to the historical average.

The current overshoot, however, is exaggerated by a long-run upward trend in depreciation, reflecting a combination of a rising capital/output ratio and a declining average life of capital goods. Margin sustainability is better judged using a net measure, excluding depreciation from both the numerator and denominator of the ratio (i.e. profits plus interest as a percentage of net product). Net margins, while high, are within the historical range and less stretched relative to the average.

The aggregate figures, moreover, obscure a bigger story: the divergence between the financial and non-financial sectors – second chart. Financial net margins are at a record high and would need to fall by 41% to restore the historical average. Non-financial margins, by contrast, are at a normal historical level, having moved temporarily below the average during the recession.

This suggests that non-financial P/E measures should not be significantly distorted by unusually-high margins currently but financial P/Es may be misleadingly low, in turn depressing market-wide valuations. The third chart shows forward P/E ratios, based on 12-month-ahead earnings estimates, for the S&P Industrials and Financial Sector indices.

The Industrials P/E is roughly in the middle of the historical range excluding the late 1990s bubble period. The Financials P/E is lower but the discount is smaller than usual. If earnings were recalculated based on average margins, moreover, the Financials P/E would be far above that of the Industrials and at the top of the historical range.

Possible margin compression, therefore, is a reason for caution on financial stocks but much less so for the rest of the market.

A prior post presented a comparison of the rebound in the Dow Industrials index from its trough in March 2009 with recoveries after the six largest twentieth-century bear markets, excluding the devastating 1929-32 decline. These bears involved index falls of 45-52%, similar to the 54% drop over October 2007-March 2009. (Prices slumped by 89% over September 1929-July 1932.)

At the time of the earlier post, the Dow had moved down to converge with the "six-bear average" of these earlier recoveries. With liquidity conditions for markets having deteriorated, a fall into the lower half of the historical range seemed likely in the short term, although a significant undershoot of the average might present a buying opportunity.

Updating the analysis, yesterday's Dow close was 8% below the six-bear mean and 1% above the bottom of the historical range – see chart.

Examining the six components, the recovery since March 2009 bears the strongest resemblance to the rebound after the January 1906-November 1907 decline. The Dow was mostly above the six-bear average during the early stages of this revival but a significant correction set in after about a year, echoing recent market weakness.

Like the recent bear, the January 1906-November 1907 decline was associated with a credit bust and financial panic that drained liquidity from markets. Both crises climaxed with the failure of a major bank – the Knickerbocker Trust Company in October 1907, Lehman Brothers in September 2008 – and a subsequent decisive "official" rescue effort (co-ordinated by J P Morgan in 1907, before the institution of the Federal Reserve). The economic consequences were similar, with a severe one-year recession in industrial output followed by a strong rebound.

Relative to the bear-market trough, the Dow is currently very close to its level at the same stage of the post-November-1907 recovery – see chart. On that occasion, prices were reaching a low and rallied by more than 20% over the following six months.

The "six-bear" evidence, therefore, suggests that market weakness will abate. Resumption of an uptrend, however, requires an improvement in liquidity indicators.

Another reason for doubting the significance of the slowdown in US private-sector payrolls in May is that business surveys have yet to signal any weakening of labour demand.

The chart shows a survey-based measure of labour demand comprising three components: the net percentage of Institute for Supply Management (ISM) manufacturing firms expanding employment, the monthly lay-off tally by outsourcing consultancy firm Challenger, Gray & Christmas and the net percentage of National Federation of Independent Business (NFIB) small firms planning to hire. Reflecting increases in the ISM and NFIB components, this composite measure rose further in May to its highest level since January 2007.

Friday's US employment numbers for May were reported as being disappointing because of a smaller-than-expected gain in private-sector payrolls. A revival in private employment incomes is a necessary component of a sustainable economic recovery.

The reaction, however, looks exaggerated, for three reasons. First, the small May rise followed strong gains in March and April, so the three-month increase remains solid at 0.4%, or 1.6% annualised.

Secondly, employees worked longer hours on average in May, compensating for the smaller jobs increase. Aggregate private-sector hours have risen by an unusually-strong 1.3%, or 5.4% annualised, over the last three months – see first chart.

Thirdly, an alternative measure of private-sector employee jobs derived from the monthly household survey shows a stronger recent gain – second chart. This measure is more volatile but may be a better indicator around turning points in the cycle, partly because it should pick up trends in smaller firms that are underrepresented in the payrolls survey.