Podcast

Important Information:
How the Fed Mines Data for New Insights
Important Information:
Nicolas Morales, Horacio Sapriza, and Chen Yeh take listeners behind the scenes on how they work with detailed, confidential datasets about businesses and individuals. They also discuss how they gain insights on topics like immigration, credit markets, and labor productivity while safeguarding privacy rights. Morales is an economist, Sapriza is a senior economist and policy advisor, and Yeh is a senior economist at the Federal Reserve Bank of Richmond.
Transcript
Tim Sablik: The Fed often describes itself as data dependent, meaning that it uses data to constantly update its picture of the economy and set monetary policy appropriately. On today's show, we're going to be talking about some of the data that Fed economists use in that process.
My guests are Nicolas Morales, Horacio Sapriza, and Chen Yeh, all at the Richmond Fed. Gentlemen, thank you for joining me today.
Nicolas Morales: Happy to be here.
Horacio Sapriza: Thank you for having us, Tim.
Chen Yeh: It's always a pleasure to be here.
Sablik: Before I joined the Fed, I had this impression that the central bank had access to all the economic data in the world. Economists just needed to pick a topic to work on and the data was right there waiting for them.
Maybe some of our listeners had the same thought, or maybe it was just me. But I know now that it's definitely not so simple. The Fed does have access to a lot of data. But, in many cases, gathering and analyzing that data to answer really difficult questions about the economy takes a lot of hard work. I invited all of you onto the show today because you're working with some interesting datasets.
To start off, could you start by introducing your research projects and the data you're working with? Nicolas, how about you go first?
Morales: Much of my research focuses on understanding how immigration — and particularly the immigration of college graduates — affects the U.S. economy. By the U.S. economy, I mean both U.S. firms and native workers in the U.S.
To understand this issue, it's important to have data on the economic outcomes of firms that hire the immigrants, such as their total employment and their revenues, as well as labor market information on the workers, such as how many employees of a firm are natives or immigrants, what are the wages the employees of a firm earn, and also how workers move across firms.
The Longitudinal Employer-Household Dynamics — or, in short, the LEHD — from the U.S. Census gives confidential access to the universe of employers and employees for generally 25 U.S. states. In that data, we can see whether workers employed by a firm are U.S. born or foreign born. We can also see which firms they work for and also their earnings.
In these projects that I'm working on, we complement the LEHD with another dataset called the Longitudinal Business [Database] — or, in short, the LBD datasets — that includes information at the firm level on the universe of employers in the U.S. It includes their total employment, their revenues, and also the industry they belong to.
As you might imagine, this data has restricted access to ensure confidentiality of the respondents. So, to publish the results of our research, we generally work together with the U.S. Census Bureau through a vigorous review process before presenting aggregate results using the data.
Sablik: Thanks. Horacio?
Sapriza: Over the years, I've worked with different datasets that basically allow me to learn about the links between financial markets and the real economy. For instance, in some studies I looked at sovereign debt and how the risk of a country's debt can affect its optimal economic policies, or how countries' recoveries relate to the maturity of the debt. For this analysis, I've used sovereign bond market pricing and volumes data, and macrodata for different countries. This data comes from both private vendors such as Bloomberg and public databases from international organizations like the IMF.
I also worked on more domestically focused studies, looking at how shocks have different effects on bank credit depending on bank characteristics. There, I combine macrodata with bank and firm loan level data from confidential sources. Of course, like Nicolas was mentioning, there are a number of controls and processes that have to be satisfied as well.
I'd also work with call reports. There, I looked at how changes in bank lending standards may impact credit to firms and households. I've used a Senior Loan Officer Opinion Survey that's collected by the Fed, providing data on the bank lending standards, terms, and demand for loans for different loan categories. So, I use a variety of datasets depending on the research project.
Sablik: Gotcha. Chen?
Yeh: My research agenda covers many topics. I deal with firm dynamics: the growth and contraction patterns of firms. For this topic, I usually use establishment- and/or firm-level data to study several phenomena on their pricing or employment decisions or productivity. The gold standard is using data from the Census Bureau. So, similar to the data that Nico has just described: confidential level microdata at the establishment or firm level. There are rooms where you work in, whole research data centers. You work in a secure environment to maintain the privacy of respondents.
One of the projects I've been working on is about how establishments are changing their hiring practices, in particular how they are hiring through staffing agencies. Traditional data sources that cover the universe of employers usually focus on direct hiring, so you wouldn't observe these staffing hires. What I'm currently doing now is focusing on manufacturers and using this data product from the Census called the Annual Survey of Manufactures and the Census of Manufactures. What is kind of interesting there is that you do see some information on the staffing hires.
Another topic that I'm working on is productivity in the construction sector. What is kind of an interesting phenomenon in construction, unlike any other sector in the U.S. economy, is that its productivity — what you can do with a given amount of inputs, how much output you get from that — it's been declining over time. Even though that fact is well known, its cause is rather unknown. For this project, I'm using data from the Census Bureau, in particular the Census of Construction Industries.
Sablik: Thinking about these projects that you all are working on, what are the sort of questions about the economy that you're hoping to shed new light on?
Morales: I've been studying immigration for a long time. In the literature, there's a longstanding question on what is the economic impact of immigrants, particularly their effect in the labor market. Since firm-level data is not easily available, the literature had generally focused on what happens when there's an inflow of immigrants in a labor market and then measuring how different workers that were in that labor market respond.
This approach was missing an important component. That was how firms that employ the immigrants respond to immigration. If we think about the U.S. immigration system, particularly the immigration system for college graduates, generally firms have a very important role in selecting what immigrants come into the country. They can sponsor these workers for a visa.
A particular avenue of these visas, a very important channel, is what is called the H-1B visa. This visa program is considered to be a potential engine for U.S. productivity and growth since it allows access to skills that may not be available easily within the country. We need to understand which firms benefit the most from hiring immigrants through this program, and also what is the impact of this program on the labor market outcome of native workers.
Sablik: Horacio?
Sapriza: I am currently working on a couple of topics. The first would be a project where I try to tease out the role of credit market sentiment on macroeconomic outcomes. In other work, we're looking at changes in bank lending standards that may affect firms' and households' access to credit and how that transmits to the broader economy. In a third line of recent work, we're looking at the extent to which shocks to the value of the collateral that firms pledge when borrowing affect their capacity to do so.
Yeh: In terms of the staffing agencies project, there has been a lot of anecdotal evidence that it has been on the rise. That, in and of itself, is very interesting. But I would say that for me, as a macro economist, it's all about productivity.
We would assume that in a productive economy, there's a lot of turnover or churn of jobs. But what we've observed the past few decades is that this job reallocation or this turnover of jobs has declined a lot in the U.S. economy. These facts on job reallocation usually pertain to direct hiring only. This is derived from IRS tax data. There's a lot of hiring or a lot of job flows that we don't see because that's happening through staffing agencies. Now we do have some data on it, and then the question is, how much of these job flows that we're supposedly not measuring could help in explaining away the decline in turnover.
And then the construction productivity project, I would say, is about productivity, obviously. But I would say the bigger picture is that U.S. productivity growth has been in a slump, especially since the early 2000s. There have been some explanations around, but we don't really understand that deeply.
Sablik: So, it seems like a lot of these projects are about trying to get more information about things that are maybe difficult to measure or answering questions that have arisen from certain macroeconomic datasets that we want to learn more about. What sort of challenges did you all face when it came to finding the right data to answer these questions?
Morales: I think the key limitation that many researchers have found in the past is really finding the right data to answer these questions. The reason is that many countries have a variety of administrative datasets that are similar to the ones we use with the U.S. Census, but most data doesn't have an indicator, for example, on whether the employees hired by a firm are immigrants or natives.
A second reason is that when you're doing an analysis where you're trying to understand which firms benefit the most for accessing immigrants or some other type of policy, generally you need to start cutting the data into different subgroups to understand whether small firms respond more than big firms and whether firms that have high productivity may respond more than low productivity firms.
You also need a dataset that has sufficient number of observations that you can break down the analysis enough to be able to distinguish the effects for different groups.
Sablik: In addition to advancing economists' understandings about these various topics — and also, I imagine, being personally interesting to each one of you — how is this research contributing to informing the policymaking process at the Fed?
Morales: From the Fed's perspective, immigration is a key contributor for population growth, and population growth generally translates also into employment or workforce growth. Immigration also has the potential to give employers in the U.S. access to skills that might not be available in the local population.
Sablik: Horacio?
Sapriza: My research contributes to policymaking by helping understand how monetary policy and other sources of fluctuations affect the real economy through financial markets and financial institutions. For instance, my work is used to provide insight on retransmission channels and exposures of financial institutions to different types of risks, domestic and foreign.
Sablik: Chen?
Yeh: Staffing is considered to be a "non-traditional" form of hiring. So, obviously, it affects employment. As we know, maintaining maximum employment is one of the mandates for the Federal Reserve.
Just to give you a concrete example, think about the employment fluctuations over the business cycle. What I found in the data is that staffing hiring is actually more prevalent directly after recession. So why does that matter? Let's say in the data you see that direct hiring is sluggish after a recession. That might not look as good, but maybe the economy is actually not doing so bad because companies are attracting workers, but not necessarily through direct hiring, but rather through staffing agencies. So, in that sense, staffing might imply that economic recoveries are not as slow as we might think.
Sablik: Well, Nicolas, Horacio, and Chen, thank you so much for joining me today and giving us all a behind-the-scenes look at the research process.