Economics 9/8/10: Ireland's Construction PMIs

This morning brought with it another bunch of wonderfully optimistic statements from the Irish 'experts' on business cycles.

Let's take in the facts:

• Ulster Banks’ Irish Construction PMI data released today showed moderating decline in Irish construction activity in July. PMI increased modestly from 44.9 in June to 45.0 in July which still means a contraction in activity.
  
• However, at 45.0 the 'improvement' in terms of slower rate of decline is within margin of error, at least one based on time series residuals (Ulster Bank won't tell us what the real underlying margin of error in PMI surveys for the sector is).
  
• So on the surface, contraction in activity is now "the slowest in three years". Which of course is only a natural statistical property - after 3 years of destruction raging across the sector, you'd get an asymptotic curve to 'stabilization', aka the bottom. This has absolutely nothing to do with any pending improvements.
  
• Residential sub-sector was the weakest, showing accelerating drop-off to 40.8 in July, from 45.4 in June. So housing continues to fall off the cliff.
  
• Commercial and civil engineering sub-sectors posted an 'improvement' in July - with the rate of collapse slowing from 45.8 to 46.0 (another statistically insignificant change) and to 43.6 form 38.4 respectively (clearly a statistically significant number). Again - the 'good news' here is a slowdown in the rate of the fall off, no real improvement.
  

The real spin stuff was, actually, in the interpretations concerning future expectations: "Future sentiment remained strongly positive in July, and improved slightly since the previous month, as over 40% of respondents expect activity to be higher in twelve months’ time."

You see, should the question have been 'Do ou expect any improvement in activity 10 years from now?' the 'improved' sentiment would have probably been even stronger.

Virtually identical analysis was presented by the Ulster Bank itself (here). Ulster Bank chief economist Simon Barry told the Irish Times that "index showed that conditions in the Irish construction sector remained “very tough”, with firms continuing to cut back sharply on their employment levels... [But] 'Looking forward, the July survey picked up a further improvement in confidence among Irish construction firms,' Mr Barry said. The rise in new business would provide “added encouragement”, he noted... 'As heartening as this development is, the increase is very modest indeed and it is probably more an indication of possible stabilisation in the sector at very weak levels rather than a strong recovery anytime soon.'"

This type of interpretation omits a very simple economic reality: after 38 months of contraction, the firms still remaining standing in for the survey are those that survived so far into the downturn. These same firms might have higher expectation of surviving into the near future as well. In other words, the entire PMI survey component suffers from survivorship bias. This bias may (or may not) be significant for several reasons:

1. Surviving firms might be biased on the optimism side because they expect to pick up a greater share of future public spending on construction due to declined competition. In other words, survivors might be looking forward to having an increased market share of a shrinking economic pie. Surely that wouldn't be indicative of 'stabilization'.
2. Surviving firms might also be collectively biased in their responses to the survey, if they have individual incentives to do so. For example, a number of Irish construction firms are currently under continued pressure from their banks. If each one of those firms were to make a signal to their lenders that 'things are going to improve soon, just wait a little longer', the resulting bias can be significant enough to induce higher optimism readings on the survey side. This is a significant enough effect in other sectors using surveys of expected future conditions to invalidate entire indices. One classical example involves surveys of expectations for future direction in Forex markets.
3. Surviving firms might also be selected on the basis of their actual exposure to the Irish market. For example - two leading surviving firms in the Irish construction sector are Kingspan and CRH. CRH derives only 4% of its revenue from Ireland and Kingspan's share of revenue accruing to Ireland is 7%. If firms are indeed selected into survivors group by their lower exposure to the Irish market, the question is then whether the expectations data they report is purely based on their perception of future trading conditions in Ireland or whether it is 'contaminated' by their reading of other markets.
   

What (1) and (3) above really suggest is that before we engage in interpreting the future expectations we need to rigorously check for a number of classical biases that might be present in the data. Only economists unaware of the hazards of interpreting survey based gauges of expectations would make the basic mistake of taking the number at their face value and interpreting them directly.

Alternatively, for a more crude correction, the survey results should not be interpreted independent of the quantitative data from contemporaneous PMI reading. In other words, one can make a conclusion that 'It is likely that in the near term there will be improvements in trading conditions in the sector' only if there are some contemporaneous signals of improvements and only if these signals are statistically strong enough.

This, of course is hardly the case, given that PMIs contemporaneous reading increased by just 0.1 from 44.9 to 45.0 - an increase that appears to be well within the margin of error.