News / Features

Newsroom

 

When an Economic Hurricane Hits Without Warning

Forecaster Arthur Kordon, The Dow Chemical Co., recounts his experience and out-of-the-box thinking following the biggest recession since the Great Depression.

How do we know when a hurricane is coming? Fortunately weather forecasting technology today allows us to see one coming days in advance. Impacted areas take precautions whether it’s general awareness, sending formal warnings to citizens or evacuating. But what happens when an economic hurricane hits without warning, such as the 2008 recession that was comparable to the Great Depression? The benefits of forecasting this type of crisis surely come to fruition.

Arthur Kordon, Advanced Analytics Leader in the Advanced Analytics Group of The Dow Chemical Co., talks about his experience when the economic crisis hit at the end of 2008, including how his group’s forecasting capabilities suddenly became top priority to key business decision makers.

Why did the recent recession take The Dow Chemical Co. – and most other businesses in the world – by surprise?
Kordon: We hadn’t experienced a strong recession in decades. In the 1970s and 80s, there were frequent (comparatively) mild recessions, and then nothing hit until the 2001 recession, which was still relatively minor. Recessions became somewhat of a secondary issue; optimism was very high and the company was growing globally.

Then the 2008 crisis hit us like a hurricane. There were record low rates in the manufacturing market in January 2009. And since Dow’s 5,000 diverse products reside in various sectors from retail to wholesale to manufacturing, we were majorly impacted.

What was your group tasked to do after the crisis?
Kordon: After receiving an urgent phone call from the Strategy Group, we were challenged to predict when these sectors would come out of the trough so that upper management could invest wisely and choose the best sectors for taking risks. This all involves determining the lag time between sectors as each one heavily affects the other.

We took three approaches: cycle characterization, sector delay characterization and turns characterization. Within each we used a combination of methods and indexes that are more or less common knowledge – developed in the early ’90s – in our field. But to apply them to our economic sectors for forecasting? That was truly out of the box for Dow.

Another component here is that we needed to deliver these forecasts within three weeks. My R&D background afforded me to usually work with a set, funded project where we thoroughly planned, tested hypotheses, replicated and proved concepts. But there was no time for that. We had to act immediately. And when you’re backed into a corner like that, you begin to think more creatively.

Tell me more about the methods you used.
Kordon: For our cycle characterization study, we applied the Maccini method that compared business cycles from sector to sector in terms of length of cycle – all based on economic activity, something like a GDP. Now our data only went back to 1992 so there had been only one official recession during that time, in 2001. So most of the cycles measured were not recessionary.

We also used the Harding and Pagan method here, which dissected the cycles in more detail. For example, we could calculate several characteristics of a cycle, such as mean durations and amplitudes during expansion and contraction.

From these methods, we found some sectors had longer or shorter cycles than others, which was good to know. And those conclusions made sense with sales and marketing projections.

For sector delay characterization, we took the classic cross-correlation function approach that correlates the lags in time between two sector time series, using analytics software to determine the lags. Additionally, we took the system identification approach, which is based on process control theory, to compare continuous and discrete models.

For our turns characterization research, we used a few valuable indexes. The Chicago Federal National Activity Index (CFNAI) gave us an idea of economic indicators to identify business cycle turning points. We also studied the Economic Cycle Research Institute Index that provides insight on leading indicators on a global scale. It’s really effective to look at a regional trend but even better to look at it globally. After all we’re a global economy now and must take different areas of the world into account. Lastly, Dow created its own version of the CFNAI, which defined indicators specific to each business unit.

Companies should watch indexes like these and signal colleagues should their sector hit, say zero, adequately warning them that a tough market may be coming.

In all, the methods and indexes gave us some asymmetrical and symmetrical insights into what markets would move when, but there wasn’t one specific approach that gave us the answer. We had to use a combination of estimates to make our best guess on the sector dynamics.

What happened when you delivered the forecasts?
Kordon: Management took our estimates very seriously. And this is a key part of it: Groups like ours can forecast all day but if decision makers are not listening and receiving the warning bells, they will fail to take necessary actions to prevent crises. The appreciation of and resulting decisions from forecasting conclusions are just as important as forecasting itself.

What’s next for your group?
Kordon: We have been receiving more and more requests for forecasting models since the 2008 crisis. And we’re delivering them every day. But I think eventually we’ll move from simple forecasting to a more refined decision-making process where we look at what types of decisions the business makes, and what improvements are possible when we apply forecasts to the overall process. We’re still learning the best approaches for this, but it’s where we’re headed.

Arthur Kordon, Advanced Analytics Leader in the Advanced Analytics Group of The Dow Chemical Company

Read More