US Forecast: April 29, 2021
Forecast Period

2021.2--2025.4 (19 quarters)

Data

The forecast is based on the national income and product accounts (NIPA) data that were released on April 29, 2021.

The Latest Version of the US Model

For purposes of this forecast the US model has been reestimated through 2021.1. These estimates and the complete specification of the model are presented in Appendix A: The US Model: April 29, 2021. Dummy variables for the five pandemic quarters, 2020.1--2021.1, have been used as explanatory variables in some of the estimated equations to account for the unexplained pandemic effects. A complete discussion of the January 28, 2018, version of the US model is in Macroeconometric Modeling: 2018. A few specification changes have been made for the current version of the model relative to the January 28, 2018, version. These are discussed in Appendix C: Changes to the US Model Beginning April 29, 2021. Unless otherwise noted, the flow variables in the model are presented in this memo and on the site at quarterly rates. To convert quarterly rates to annual rates, just multiply by four.

The current forecast is discussed in more detail in What Do Price Equations Say About Future Inflation? In this paper two forecasts are presented in Table 7. The first has the estimated Fed rule (equation 30) turned off and RS set to 0.05 for all quarters of the forecast period. The second uses the estimated Fed rule. In this memo the first forecast is discussed, namely with the Fed rule dropped and RS taken to be 0.05. Note that the default when solving the model on this site is to use the Fed rule. This duplicates the second forecast in Table 7. If you want to run the first forecast in Table 7, drop the RS equation, 30, and set RS to 0.05 for all quarters of the forecast period. This will duplicate the first forecast in Table 7, which is discussed in this memo. The only difference between the two forecasts is the treatment of the Fed rule. .

Assumptions Behind the Forecast

The table below gives the growth rates that were assumed for some of the key exogenous variables in the model.

         Assumed Growth Rates (annual rates)

           Forecast             
          Assumptions           

TRGHQ         see text          
COG           3.0               
JG            3.0               
TRGSQ         1.0 from 2020.1 base value
TRSHQ         4.0 from 2020.1 base value
COS           3.0               
JS            1.0
EX            3.0
PIM           3.0

The first seven variables are the main government policy variables in the model aside from tax rates. TRGHQ is real federal government transfer payments to households, COG is real federal government purchases of goods, JG is federal government civilian employment, TRGSQ is real federal government transfer payments to state and local governments, TRSHQ is real state and local government transfer payments to households, COS is real state and local government purchases of goods, JS is state and local government employment, EX is real exports, and PIM is the import price deflator.

A key issue for the current forecast is how to account for the American Rescue Plan Act (ARPA) passed in March 2021. The Congressional Budget Office (CBO) and the Joint Committee on Taxation (JCT) have estimated the budget outlays from this act: $1,088 billion in FY2021, $476 billion in FY2022, $115 billion in FY2023, and then relatively small amounts after that. Some of this was spent in 2021.1. From the NIPA release on April 29, 2021, nominal federal transfer payments to persons was larger in 2021.1 versus 2020.1, the last ``normal'' quarter before the pandemic, by $686 billion at a quarterly rate. Grants-in-aid to state and local governments was larger by $39 billion, and subsidies was larger by $82 billion. This total, $807 billion, is assumed to be due to the ARPA. This leaves $281 billion left for 2021.2 and 2021.3 using the CBO and JCT estimate of $1,088 billion for FY2021. I have allocated this 60/40 in the two quarters, so $169 billion in 2021.2 and $112 billion in 2021.3. For the next four quarters I have allocated the $476 evenly, so $119 billion each. For the next four quarters I have allocated the $115 billion evenly, so $29 billion each.

These values are in nominal terms. To convert them to real terms, I took the value of the GDP deflator in 2021.1, let it grow at an annual rate of 3 percent, and used these values to deflate the nominal values. The 11 values over the 11 quarters in billions of dollars are 145, 96,101, 100, 99, 99, 24, 24, 24, and 23. (The actual $807 billion nominal value in 2021.1 is $699 billion in real terms.) Although some of this additional spending will take the form of increased real grants in aid and increased subsidies, for purposes of the forecast all has been put in real federal government transfer payments to persons, variable TRGHQ in the model. The value for subsidies was taken to be $20 billion in each of the 11 quarters, roughly its value in 2020.1. Real grants in aid, variable TRGSQ in the model, was taken to grow at an annual rate of 3 percent from 2021.2 on using as a base value its value in 2020.1. In addition, real state and local government transfer payments to persons, variable TRSHQ in the model, was taken to grow at an annual rate of 3 percent using as a base value its value in 2020.1. (The values of TRSHQ were higher during the pandemic as state and local governments passed on some of the increased federal grants in aid to persons. Since only normal growth is assumed for real real grants in aid for the forecast, only normal growth was assumed for TRSHQ.) TRGHQ was taken to grow at an annual rate of 3 percent using as a base value its value in 2020.1 and then the additions discussed above were added to these values.

Since some of the additional spending from ARPA will go to subsidies and grants in aid, the implicit assumption used here is that the multiplier effects from these two variables are the same as the multiplier effects from TRGHQ. The real output multipliers from increasing real $TR$ by 1 for the first 11 quarters of the forecast period are respectively: 0.11, 0.25, 0.36, 0.45, 0.51, 0.55, 0.58, 0.60, 0.62, 0.63, and 0.64. The initial effects are thus small, rising to a multiplier of about half after 4 quarters. As is obvious from the large increases in the personal saving rate after the pandemic stimulus payments, households initially save much of the increase transfer payments.

Some of the other assumptions for the forecast are as follows (all growth rates are at annual rates): tax rates unchanged from their 2021.1 values, potential output (YS) growing at 3 percent, labor productivity (LAM) growing at 1.5 percent, and the relative price of housing (PSI14) growing at 5 percent. In addition, the estimated Fed rule, equation 30, is dropped and RS is assumed to be unchanged from its 2021.1 value, which is 5 basis points.

Nothing was done about possible tax and spending changes from the Biden administration's proposed infrastructure plans. The current forecast is obviously a conditional forecast, conditional on nothing new done after the ARPA. As least for the first year or two ``nothing new'' is likely not a bad approximation since it will take time for the legislation to be passed if it is and for the beginning of the increased spending and taxes.

The Forecast

Forecasts of selected variables are presented in the following: Forecasts of selected variables---html, Forecasts of selected variables---pdf file. If you want more detail, click "Solve current version" after "US Model," create a data set, and then go immediately to "Examine the results without solving the model." You can then examine any variable in the model.

Real GDP Growth and the Unemployment Rate and the Labor Market: The forecast has real GDP growing in the next four quarters at 11.9, 9.0, 5.9, and 4.0 percent, respectively. (All growth rates in this memo are at annual rates.) The unemployment rates in the four quarters are 5.5, 4.7, 4.0 and 3.6, respectively. This large growth rate in 2021.2 is in part due to household wealth, which is large from past transfer payments saved and from past large increases in stock and housing prices. This has a large effect on household expenditures, including housing investment. The high growth rate is also due in part to a large predicted inventory correction in 2021.2 (inventory investment was negative and large in absolute vlaue in 2021.1.) In addition, $TRGHQ$ is large from the ARPA. The large growth rates in the next three quarters are from the continuing wealth effects and the continuing large transfer payments. None of this is, of course, surprising. The U.S. economy has had a huge fiscal stimulus, a huge increase in financial and housing wealth, and an accommodating monetary policy.

It will be instructive to give a few more details. Comparing 2021.1 to 2019.4, private jobs (JF) fell by 7.93 million, government jobs (JG+JM+JS) fell by 1.12 million, and the number of people holding two jobs (LM) (moonlighters) fell by 1.55 million. The number of people employed (E), which is jobs minus moonlighters, thus fell by 7.50 million. Had there been no change in the labor force, the number of people unemployed (U) would have increased by 7.50 million. In fact it increased by only 4.09 million because the labor force (L1+L2+L3) fell by 3.41 million. The unemployment rate (UR) rose from 3.6 percent to 6.2 percent.

How fast is the economy forecasted to come back? Comparing the forecast values for 2022.1 to the actual values in 2021.1, private jobs rose by 5.78 million, government jobs rose by 0.24 million, and moonlighters rose by 0.97 million. The number of people employed thus rose by 5.05 million. The labor force rose by 0.96 million, so the number of people unemployed fell by 4.09 million. The unemployment rate fell from 6.2 percent to 3.6 percent. Had the labor force been forecast to come back to where it was, the fall in the unemployment would obviously been less. One of the reasons for the small forecasted rise in the labor force relative to how much it fell is that household wealth has a negative effect on labor supply in the labor force participation equations, and, as noted above, there are large increases in household wealth. The labor force is not back to its 2019.4 value until 2023.3. The number of private jobs is back by 2022.3.

Inflation: Inflation as measured by the growth of PF in the next eight quarters is 1.6, 2.5, 3.5, 4.2, 4.6, 4.8, 4.8, and 4.6, respectively. The reason for the low inflation forecasts for the first two quarters is that the unemployment rate is still fairly high. Once the unemployment rate gets down to about 3.5 percent, the inflation forecasts increase to over 4 percent. An interesting question is if this turns out to be the case, will the Fed step in and if so how effective will it be? Remember that the current forecast is conditional on the Fed not responding. The Fed rule, equation 30, has been turned off. You can easily on the site add the Fed rule back in and see how much difference this makes.

Possible Experiments to Run

This forecast assumes no bad financial shocks. The growth predictions would be worse if there are. Possible bad shocks are financial market reactions to growing inflation. Below is a discussion of how to incorporate these shocks into the model.

The assumption of no bad shocks, which is used for the forecast, means that stock prices, housing prices, and import prices grow at historically normal rates. There are no negative wealth shocks through falling stock prices and housing prices. Changes in asset prices like stock prices, housing prices, exchange rates, and oil prices are essentially unpredictable. One can use the US model to analyize the effects of asset price shocks, but the shocks themselves cannot be predicted. The best one can do in a forecast is to assume some historically average behavior of asset prices, which has been done here.

To examine the effects of asset price shocks, experiments can be run using the model in which stock prices (variable CG), housing prices (variable PSI14), and import prices (variable PIM) are changed. This allows one to examine the sensitivity of the forecast to changes in these values.

To review, oil price shocks and exchange rate shocks are handled through changes in PIM. Housing price changes are handled through changes in PSI14. Changes in PSI14 change PKH relative to PD and thus change housing wealth, PIH*KH. This affects consumption expenditures through the total wealth variable AA (equations 1, 2, and 3). It affects housing investment through the housing wealth variable AA2 (equation 4). Regarding the stock market, each change in the S&\P 500 index of 10 points is a change in CG, the capital gains variable in the model, of about $100 billion. If you think that the S&\P index will fall, say, 400 points, you should drop the equation for CG and change CG by about -$4,000 billion. See the discussion in Section 7.2 of The US Model Workbook, April 29, 2021. This will have a negative effect on real output growth because of a negative wealth effect.

Other possible experiments are to change the various exogenous government policy variables in the model. For example, you may want to change some of these variables based on what infrastructure plan you think will be passed, including possibe tax rate changes.