The MIT Master's Program in Engineering and Management
Editor's note: Felipe Bustos is a Captain in the Chilean Air Force who serves as an advisor to the CEO of Empresa Nacional de Aeronáutica (ENAER), a Chilean national aircraft maintenance, repair, and parts manufacturing company. He recently graduated from MIT's System Design and Management Program (SDM) where he earned an MS in Engineering and Management. His thesis research, conducted with SDM Fellow Fernando Barrazza, SDM '10, involved using systems thinking and publicly available data to track business cycles in the manufacturing sector.
Although manufacturing represents 12% of the US economy and the Obama administration is emphasizing manufacturing as a way to stimulate the creation of high-quality jobs, most data related to manufacturing are expensive and suffer from inherent biases. With my co-author, Fernando Barraza, I saw the opportunity for a metric that would characterize the sector as a whole, using a simple yet meaningful mathematical representation derived from public data.
Our metric is the Manufacturing Composite Index of Leading Indicators (MCI) and our methodology is based on the proven work of the US National Bureau of Economic Research. We used data from the US Census Bureau and the US Bureau of Labor Statistics which included time series for new orders, shipments, total inventory, capacity utilization, and average weekly hours of manufacturing.
We adjusted for seasonality and inflation so that we could focus on the business cycle. We used seasonally adjusted indices and the Bureau of Labor Statistics' Chained Consumer Price Index for All Urban Consumers (C-CPI-U). We then weighed the contribution of each index by using the inverse of the standard deviation to minimize the inherent randomness of each index.
After several months of intensive data mining, we compiled our first set of graphs, plotting the MCI against the US Manufacturing Gross Domestic Product (GDP) for several subsectors. We arbitrarily started with Computers and Electronic Products and Primary Metals. The former didn't show the correlation we were expecting, but the latter proved to be a great example, and the results were a milestone in our research.
When we benchmarked the MCI against the GDP, we found that the MCI anticipates fluctuations in the GDP by 5 to 9 months. We had come up with a metric that could help predict recessions. We showed that the MCI effectively describes the manufacturing sector and is a tool that can help managers and decision-makers plan for contractions and expansions in the sector.
The novelty of our work lies in its reliance on publicly available data. The US Census Bureau and the US Bureau of Labor Statistics provide highly detailed, reliable secondary data. The most relevant reports for our purposes were the Advanced Report on Durable Goods Manufacturers' Shipments, Inventories, and Orders, the Quarterly Survey of Plant Capacity Utilization, and information from the Bureaus' websites.
Our MCI method avoids the potential biases and excessive costs associated with economic indicators that use senior managers or private sources of data. There are practically no limits in terms of accessibility, utilization, or information gathering methodology with our public sources of data.
We validated the MCI method by using it to identify fluctuations in the Canadian manufacturing sector. We found that, as with the US manufacturing MCI, the Canadian MCI anticipates Canadian manufacturing activity and correlates with Canadian GDP.
The MCI correlates with GDP on 18 of 20 manufacturing subsectors defined by the North American Industrial Classification System. In looking at industry subsectors, we found that some lead their respective GDPs more than others. Food products, petroleum and coal products, and primary metals and fabricated metals are prime movers in manufacturing. They react early to changes in expectations for future demand and therefore provide advance indications of contractions and expansions.
We identified these prime movers by taking a systems perspective. We examined the different stakeholders within each subsector and how they interact and transfer value and information. This process helped us understand the dynamics within the industry. In particular, we were able to see how decisions in a specific subsector revealed unexpected behavior in other subsectors.
However, players in these prime mover subsectors do not necessarily have better information or better forecasting techniques. It's possible that this leading activity is a consequence of how the industry operates. For example, these players could simply be responding to orders from downstream manufacturers.
Although the MCI cannot directly forecast GDP, it provides meaningful information about the most critical economic events: turning points from growth to decline that anticipate recessions. In practical terms, the MCI method is useful for companies in the US. Because we can calculate the MCI monthly with a one-month lag, we're able to predict turning points before the release of official quarterly information.
Businesses can use our method as a management tool in several ways: delaying expansion or acquisition plans to wait for better prices, renegotiating contracts for raw materials, adjusting hiring plans, decreasing capacity utilization to reduce inventory, and shifting sales strategies. Investors can use the MCI method to adjust their valuations of companies.
Businesses can monitor suppliers more efficiently, for instance, by paying attention to prime mover subsectors. An automaker planning to buy raw materials for the next quarter could track the MCI for primary metals. If the primary metals MCI is declining, the automaker could delay the purchase until the primary metals subsector weakened further in hopes of getting a better price and reducing the automaker's cost of materials.
Businesses can also create Inventory Coverage indices for each subsector by combining the subsector's Total Inventory and Shipments time series and calculating the average, maximum, and minimum historical values for each Inventory Coverage index. Businesses can then derive safe and warning zones with respect to market behavior. If a manufacturer's inventory coverage falls in the warning zone, for example, the company has a good indication that it should revise its strategy.
For example, looking at the Inventory Coverage Index for machinery manufacturing shows the distinct difference in how Caterpillar and John Deere responded to the recession that began at the end of 2007. Caterpillar's inventory days rose well above the industry average while John Deere's dropped below the average.
It's ambitious to predict an economic contraction using only the MCI. Doing so disregards other important information such as income, employment, and wholesale-retail sales. Nevertheless, the MCI contains enough information to anticipate a decline. Add diffusion indices which measure how widespread a trend is and it's possible to anticipate the depth of an impending downturn.
The MCI would have indicated the recessions of 2001 and 2008 well before the official government declarations. The 2001 recession began in March and ended in November of that year, and the official declaration of the recession was issued on November 26. The MCI would have given its first indication of the recession in March. Likewise, the MCI would have identified the 2008 recession, which began in December 2007 and ended in June 2009, in October 2008. The official declaration came on December 11, 2008.
It's possible to create many different manufacturing-related metrics depending on the datasets gathered for the MCI. One possibility is the Inventory Coverage Indicator, which provides a view of the relationship between inventory and sales in a subsector. Companies can then compare their performance with its subsector as a whole.
The MCI can be a better indicator for the larger economy if time series data outside of manufacturing is added. Leading indexes from other sectors such as stock prices, new construction permits, and money supply can incorporate prime movers that normally precede manufacturing.
The data for our project span 1999 through 2011. Expanding the timeframe will provide a better calibration of the MCI. We expect that continued monitoring of the MCI will be important for economics, business, and policy making. It should be particularly important for smaller companies that can't afford to hire consultants.
To that end, our thesis advisor, Prof. David Simchi-Levi, is considering creating a quarterly or monthly report on US manufacturing based on the MCI. We will be especially pleased if the MCI is made freely available through an MIT website.