Economy / Other

How much can policymakers rely on performance metrics?

Most of us nowadays use Yelp, Urbanspoon, or something like those platforms to explore a restaurant. When browsing the Yelp website to find a good restaurant, I first filter out the neighborhoods and category, and then pick several restaurants based on star ratings, dollar signs, and number of reviews. Then I put a little extra effort to read through the reviews to check whether it has outdoor seating or to hear about balanced opinion about people’s personal experiences. Sometimes, I have a pleasant dining experience in places with relatively lower star ratings, and vice versa. This is an example of how we use metrics to make small decisions in our lives. Numbers are partially reliable in describing people’s experiences and they are unable to tell the whole story.

Federal agencies release numerous reports that measure the performance of the economy on a regular basis. BLS releases monthly employment reports that reflects the labor market conditions.  The Census Bureau releases population, demographic, and housing unit estimates every year or five years. The BEA of the Department of Commerce releases quarterly GDP reports, a comprehensive index of the country’s economic health. The BEA and Census Bureau co-release monthly trade reports. These economic indicators provide meaningful information but again, they have limitations in capturing the full economy. Sometimes, the national statistics are not up to date or they are missing important information.

For example, according to a recent report from the National Bureau of Economic Research, there are astonishing examples of the potential for Artificial Intelligence to greatly increase productivity and economic welfare, but national statistics fail to capture this. One application of AI is machine learning technologies, which improve themselves over time, and transform machines to perform some basic types of perception. Another one is cloud computing, where machines can share knowledge and skills almost instantaneously with others, thereby increasing the amount of data that any given machine learner can use. Using these methods, machines have made substantial gains in perception and cognition. More examples and details about improvements of the AI can be found in the report. However, the productivity benefits of it are not reflected in current national metrics. For instance, aggregate labor productivity growth in the U.S. averaged 1.3% per year from 2005 to 2016, less than half of the 2.8% annual growth rate sustained over 1995 to 2004. The report explains that one of the explanations for such discrepancy is that “economic statistics are not up to the task of accurately measuring the benefits already being enjoyed by the new wave of technologies.” For example, smartphones, online social networks, or downloadable media deliver significant utility even if they account for a small share of GDP due to their low price. Moreover, AI capital is difficult to measure because it is intangible and thus is difficult to quantify.

Gaps in national statistics appear in housing statistics as well. According to a recent blog from the Brookings Institution, policymakers across the U.S. would like to encourage more housing development to address tight housing inventory, but do not have an accurate metric of how much housing their communities need. Housing datasets released by Federal statistical agencies do not assess the balance between housing supply and demand in a timely and easily interpreted manner. The blog presents several challenges to developing such a metric with existing data. One of the explanations is that housing is both a stock and a flow, but most datasets capture only one dynamic. Another one is that we mostly rely on counts of new construction but there are many other ways that supply adjusts.

To conclude, Federal governments bear responsibility to provide clear and accurate metrics as policymakers rely on national statistics to construct effective policy solutions. But constantly improving and updating them is costly and challenging. Furthermore, numbers themselves, no matter how accurately and timely updated, are sometimes inadequate in understanding the full scope of an issue. Therefore, policymakers should put extra effort going through supplementary information such as literature or raw data to fill in missing gaps of the current Federal statistics.