| An Analysis of the Clinton and Dole Economic Plans |
| Ray C. Fair, September 4,
1996 Introduction An important use of a model like the US model is to examine the effects of fiscal-policy changes on the macro economy. At the moment there are three major federal government budget proposals for the period from now through fiscal year 2002: 1) the latest Congressional budget resolution, 2) President Clinton's budget proposal, and 3) the budget proposal of the Dole campaign. This note will use the model to analyze the macro implications of these proposals. You should read Chapters 2 and 5 of The US Model Workbook before reading this note. Advanced users may also want to read Chapter 11 of Fair (1994), which provides a detailed discussion of the policy properties of the model. I have tried to be clear in this note on the assumptions I made for each experiment. If you disagree with some assumptions, it should be clear which variables you need to change to modify the assumptions. Running alternative experiments from those in this note is a good way of examining the sensitivity of the results to alternative views and is also a good teaching device. The period considered is from now through fiscal year 2002: 1996:3-2002:3. Unless otherwise noted, "year" means fiscal year in this note. Also, the growth rate of a variable for 1996:2 over 1993:3 will be referred to as the growth rate "over the last three years" even though the period is only 11 quarters in length. This is the growth rate over the last three fiscal years excluding 1996:3. All growth rates are quoted at annual rates in this note. You should be familiar with the following government variables in the model: Federal Government:
COG Purchases of goods (real)
PG Price index for COG, so PG*COG is purchases of goods (nominal)
(PG is endogenous)
JG Number of civilian government jobs
PUG-IGZ Consumption expenditures (PUG is endogenous)
TRGH Transfer payments to persons
TRGS Grants-in-aid to State and Local governments
SGP Surplus or deficit (-), national income and product accounts
(SGP is endogenous)
D1G Personal income tax parameter
TAUG Estimated progressivity personal income tax parameter
THG Level of personal income taxes (THG is endogenous)
D2G Corporate income tax parameter
State and Local Governments
COS Purchases of goods (real)
JS Number of government jobs
TRSH Transfer payments to persons
SSP Surplus or deficit (-), national income and product accounts
(SSP is endogenous)
D1S Personal income tax parameter
Before proceeding to an analysis of the economic plans, a brief discussion is given of the differences between forecast runs and policy runs. This is followed by discussion of a change that was made in estimating the progressivity tax parameter TAUG. Then the discussion of the plans begins. Forecast Runs Versus Policy Runs When I make an "official" forecast with the model, which I do once a quarter, I need to make predictions of the future values of all the exogenous variables in the model, including the exogenous government policy variables. When I make predictions of exogenous federal government policy variables, I do not always assume that the currently existing federal budget resolution will be carried out. In the Gramm-Rudman days, for example, the government consistently deviated from its announced plans, and for forecasting purposes it would not have been sensible to use the "planned" government values. I usually assume that the future growth rate of a government policy variable will be close to its growth rate in the past two to four years. This "business as usual" assumption may differ sharply from what the government says it is going to do. When doing policy analysis instead of pure forecasting, on the other hand, it may be of interest to take the government at its word and use the planned values. For example, one may be interested in the question of what the macro economy would be like if the plans were carried out. For the experiments below I have used the planned values, and, as will be seen, some of these values differ sharply from those used for the latest forecast (August 2, 1996). Finally, mention needs to be made of how the exogenous state and local government variables are treated. For both forecast and policy runs I make assumptions about these variables that imply that the surplus of state and local governments, SSP, is roughly unchanged over time. In practice SSP does not change much over time, and so it does not seem sensible to choose state and local government policy options that imply large changes in SSP. Estimating TAUG, the Federal Progressivity Tax Parameter The federal tax code has been and currently is somewhat progressive,
and it is
complex. In a model it is necessary to try to capture the tax code with a
few parameters,
and estimating TAUG is an attempt to approximate the progressivity in the
tax code. The
procedure for estimating TAUG is discussed in Fair (1994), pp. 45-46.
THG/POP is regressed
on a constant, DUMi(YT/POP), i = 1,...,27, and (YT/POP)*(YT/POP), where
THG is the level
of federal personal income taxes, POP is the level of population 16 and
over, YT is the
level of taxable income, and DUMi is a dummy variable that takes on a
value of one for
quarters within subperiod i and zero otherwise. The 27 subperiods are
assumed to be
periods in which there were no major changes in the tax code. The
estimate of the
coefficient of (YT/POP)*(YT/POP) in this regression is TAUG. Once TAUG is
determined, D1G
is defined to be THG/YT-(TAUG*YT)/POP. The personal income tax equation
in the
model---equation 47--- is then While working on this note, I discovered I had done something foolish regarding the estimation of TAUG for the latest forecast. The last subperiod used in Fair (1994) is 1992:1-1993:2, and for the forecast I just extended this subperiod to 1992:1-1996:2. This treatment failed to account for the changes in the tax code that were passed in late 1993. Also, 1996:2 was an unusual quarter regarding the timing of tax collections (see the Survey of Current Business, August 1996, p. 6), and THG was unusually high in this quarter, which I did not account for. Failing to account for these changes led to an estimate of TAUG that is probably too high. To check this, I replaced the last two subperiods, 1991:1-1991:4 and 1992:1-1996:2, with three others, 1991:1-1993:4, 1994:1-1996:1, and 1996:2-1996:2. I picked these last three subperiods (as I had done earlier for the others) by examining a plot of THG/YT and observing where there were large changes. The new regression led to a value of TAUG that was about half the size of the original one (.005245 versus .010846), and so the original one is probably too high. The main consequence of using a too-high value of TAUG for the forecast is that federal government personal income tax collections are too high (and thus the deficit too low), especially after two or three years. For purposes of the experiments in this note, I have used the new value of TAUG (.005245). Also, I reestimated the entire model using the new value of TAUG so that the coefficient estimates would be completely consistent with the data. For the next official forecast I will use the expanded set of subperiods in the estimation of TAUG and in general will try to be more careful. The Congressional and Clinton Budget Plans We are now ready to begin the policy analysis. This will be done in steps. The first step is to examine the Congressional and Clinton budget plans and see what they imply about the values of the exogenous policy variables in the model. The second step is to solve the model using these values and examine the results. The results are the model's estimate of what the economy would be like if the plans were carried out. The third step is to examine Dole's budget plan and see what it implies about the values of the exogenous policy variables. Finally, the fourth step is to solve the model using the Dole values and examine the results. This may be surprising to some, but at least from a macro perspective the Congressional and Clinton budget plans are quite close to each other. Consider first the Congressional budget plan, which is reported in the August 1996 CBO report, Chapter 2, Table 5. The cuts in "mandatory" spending from the previous baseline for the six fiscal years, 1997-2002, are, respectively, 19, 27, 46, 63, 81, and 109 billion dollars. The cuts in "discretionary" spending are 11, 21, 39, 54, 80, and 100 billion dollars. There are also some modest tax cuts proposed: 17, 18, 21, 21, 20, and 15 billion dollars. All these changes are from the baseline reported in Table D-2 of the May 1996 CBO report. Now consider the Clinton budget plan, which is reported in Table 15 of a report on the 1997 budget (http://www.doc.gov/BudgetFY97/msr_sum.html). The mandatory cuts in the six fiscal years are 12.0, 17.9, 32.9, 45.4, 61.4, and 95.2 billion dollars, and the discretionary cuts are 4.6, 6.5, 26.2, 41.8, 68.1, and 83.5 billion dollars. In addition, there are modest corporate tax increases, 6.4, 8.4, 10.2, 10.7, 10.9, and 13.1 billion dollars, and modest personal income tax decreases, 13.1, 16.6, 19.1, 24.4, 20.0, and 5.4 billion dollars. Although the Clinton spending cuts are slightly lower and the net tax cuts slightly lower, the big picture is clearly the same between Clinton and Congress. The following table may help put these numbers in perspective. For this table (and for all the policy experiments in this note) I have assumed that the mandatory cuts come out of TRGH, federal transfer payments to persons, and that all discretionary cuts come out of PUG-IGZ, federal consumption expenditures. TRGH: (billions of current dollars)
1996 1997 1998 1999 2000 2001 2002
May 1996 CBO baseline 742 789 837 886 937 992 1050
Congressional cuts -19 -27 -46 -63 -81 -109
New Congressional baseline 770 810 840 874 911 941
Clinton cuts (rounded) -12 -18 -33 -45 -61 -95
New Clinton baseline 777 819 853 892 931 955
Actual growth rate (1996:2-1993:3) 5.4%
Actual growth rate (1996:2-1995:2) 5.8%
May 1996 CBO baseline growth rate (2002-1996) 6.0%
New Congressional baseline growth rate (2002-1996) 4.0%
New Clinton baseline growth rate (2002-1996) 4.3%
PUG-IGZ: (billions of current dollars)
1996 1997 1998 1999 2000 2001 2002
May 1996 CBO baseline 455 470 480 496 514 524 544
Congressional cuts -11 -21 -39 -54 -80 -100
New Congressional baseline 459 459 457 460 444 444
Clinton cuts (rounded) -5 -7 -26 -42 -68 -84
New Clinton baseline 465 473 470 472 456 460
Actual growth rate (1996:2-1993:3) 0.7%
Actual growth rate (1996:2-1995:2) 1.5%
May 1996 CBO baseline growth rate (2002-1996) 3.0%
New Congressional baseline growth rate (2002-1996) -0.4%
New Clinton baseline growth rate (2002-1996) 0.2%
Two things are clear from the above numbers. First, as just noted, the Congressional and Clinton plans are close. Congress has a growth rate of 4.0 percent for TRGH and -0.4 percent for PUG-IGZ, and Clinton has 4.3 and 0.2. Second, these growth rates imply large changes in spending patterns, compared to both the May 1996 CBO baseline and recent growth rates. It should be noted that the May 1996 CBO baseline for PUG-IGZ that I am using does not incorporate an item in Table D-2 called "Required Reductions in Discretionary Spending," which are "Unspecified reductions needed to reach CBO's capped baseline." These reductions in the six fiscal years are 0, 12, 12, 14, 8, and 11 billion dollars. I did not incorporate these reductions because they are so vague, but an alternative experiment would be to subtract them from the May 1996 CBO baseline for PUG-IGZ above and proceed from this new baseline. Exogenous Values Used for the Congressional and Clinton Plans What would the economy look like if the Congressional or Clinton plan was carried out? The following are assumptions I used for the two plans. (Remember that the forecast period is 1996:3-2002:3.) Growth Rates
Congressional Clinton August 2, 1996 Actual
Plan Plan Forecast 1996:2-1993:3
TRGH 4.0 4.3 6.0 5.4
COG -5.6 -4.8 0.0 -1.2
JG -5.6 -4.8 0.0 -1.7
TRGS 5.8 5.8 9.0 5.9
TRSH 6.0 6.0 9.0/11.0 10.9
COS 1.0 1.0 0.0/-6.0 4.2
JS 1.0 1.0 1.5 1.6
D1S 1.5 1.5 0.0 -2.1
Notes: 9.0/11.0 means 9.0% through 1998:4 and 11.0% thereafter.
0.0/-6.0 means 0.0% through 1999:4 and -6.0% thereafter.
In addition:
Congressional plan: D1G was cut by .00304 beginning 1996:4.
Clinton plan: D1G was also cut by .00304 beginning 1996:4 and D2G was
increased by .0153.
The Congressional experiment consists of using the above Congressional-Plan values in place of the August 2, 1996, values. In addition, the new value of TAUG was used, along with the coefficient estimates that correspond to this value. The value for D1G for 1996:3 was taken to be the 1996:1 value (not the unusually high 1996:2 value), and the values for D1G for 1996:4 and beyond were taken to be the 1996:1 value less .00304. Otherwise, everything else was the same as it was for the August 2, 1996, forecast. The Clinton experiment is the same as the Congressional experiment except for the values for TRGH, COG, JG, and D2G. The value for D2G for 1996:3 was taken to be the 1996:2 value, and the values for D2G for 1996:4 and beyond were taken to be the 1996:2 value plus .0153. The decrease in D1G corresponds to a tax cut in fiscal 1997 of roughly 17 billion dollars and then to slightly higher tax cuts beyond that as taxable income rises. The same cut was used for both the Congressional and Clinton plans because the proposed personal income tax cuts are so close. Although the Clinton plan eliminates the tax cut in fiscal 2002 if the deficit target is not met, this option is not pertinent because, as will be seen, the deficit target is met. The increase in D2G, the corporate income tax parameter, for the Clinton plan corresponds to a tax increase in fiscal 1997 of roughly 8 billion dollars and then to slightly higher tax increases beyond that as corporate profits rise. The choice of values for the state and local governments needs some explanation. As discussed above, these values should be chosen so as to keep SSP, the state and local government surplus, roughly unchanged over time. For the August 2 forecast I let TRSH, state and local government transfer payments to households, grow at 9 percent through 1998:4 and then at 11 percent thereafter, which is in line with the actual growth rate for the last three years of 10.9 percent. In order to plug the hole in state and local government budgets caused by this large growth in transfers, I allowed TRGS, federal grants in aid, to grow at 9 percent, considerably higher than the actual growth rate of 5.9 percent for the last three years. Also, COS, state and local government purchases of goods, was assumed to be flat through 1999:4 and then fall at a 6 percent rate after that, which is considerably lower than the 4.2 percent actual rate for the last three years. There are clearly many other choices that could have been made to keep SSP roughly unchanged. Transfer payments could have been assumed to grow more slowly, tax rates could have been increased, government employment (JS) could have been assumed to grow less, etc. From a macro perspective it does not matter too much how SSP is kept in line as long as it is somehow. For the Congressional and Clinton experiments a different choice of values was made for the state and local government variables. The CBO baseline path has TRGS, federal grants in aid, growing at 5.8 percent, and I have used this number for the two experiments. This means there is less federal money going to the state and local governments than was assumed for the August 2, 1996, forecast. To keep SSP roughly unchanged, I assumed the following: TRSH growing at 6 percent (compared to 10.9 percent in the last three years), COS growing at 1.0 percent (compared to 4.2 percent in the last three years), JS growing at 1.0 percent (compared to 1.6 percent in the last three years), and the personal income tax parameter D1S growing at 1.5 percent (compared to -2.1 percent in the last three years.) SSP is thus kept roughly unchanged mostly by having spending grow less than in the past three years, especially TRSH, combined with a modest personal income tax increase. Again, it would not matter too much if other assumptions were used as long as SSP remained roughly unchanged. Results for the Congressional and Clinton Experiments The results of solving the model for the above Congressional assumptions are in dataset BASECONG with a password of BASECONG, and the results of solving the model for the Clinton assumptions are in BASECLIN with a password of BASECLIN. These datasets are like BASE. You can copy them to your own dataset and then examine the results in detail. You can also use your dataset to make different assumptions from what I have done and run a new forecast. This is where I hope you have fun. If you do not agree with how I have interpreted the plans, you can easily run alternatives. Also, if you are happy with my assumptions but don't like some of the model's properties, you can modify the model (e.g., change coefficients or use add factors) and run alternative forecasts. A few selected values from the two experiments are as follows: Congressional Experiment Clinton
Experiment
% % % %
GDPR GDPR GDPD UR RS -SGP SSP GDPR GDPR GDPD UR RS -SGP SSP
1996:3 6933 2.8 1.7 5.4 5.3 155 101 6934 2.9 1.7 5.4 5.3 156 101
1996:4 6970 2.2 1.9 5.5 5.5 168 100 6972 2.2 1.9 5.4 5.5 163 100
1997:1 7005 2.0 1.9 5.6 5.6 164 99 7008 2.1 1.9 5.6 5.6 160 99
1997:2 7034 1.7 2.0 5.7 5.6 160 97 7039 1.8 2.0 5.7 5.6 157 98
1997:3 7061 1.5 2.0 5.9 5.5 157 95 7067 1.6 2.0 5.9 5.6 155 96
1997:4 7088 1.6 2.0 6.1 5.5 154 93 7095 1.6 2.1 6.1 5.5 153 94
1998:1 7117 1.6 2.1 6.3 5.4 150 90 7125 1.7 2.1 6.3 5.5 150 92
1998:2 7147 1.7 2.1 6.5 5.3 146 88 7157 1.8 2.1 6.4 5.4 147 90
1998:3 7181 1.9 2.1 6.6 5.2 139 86 7192 2.0 2.2 6.5 5.3 142 88
1999:4 7218 2.1 2.2 6.7 5.1 131 85 7230 2.1 2.2 6.6 5.2 135 87
1999:1 7258 2.3 2.2 6.8 5.0 121 84 7271 2.3 2.3 6.7 5.1 126 86
1999:2 7302 2.4 2.3 6.8 4.9 109 84 7315 2.5 2.3 6.7 5.1 115 86
1999:3 7347 2.5 2.3 6.9 4.9 97 84 7362 2.6 2.4 6.8 5.0 104 86
1999:4 7394 2.6 2.4 6.9 4.8 84 84 7409 2.6 2.5 6.8 5.0 93 87
2000:1 7442 2.6 2.5 6.9 4.8 70 85 7458 2.6 2.5 6.8 5.0 81 87
2000:2 7490 2.6 2.6 6.9 4.7 57 85 7507 2.7 2.6 6.8 4.9 69 88
2000:3 7539 2.6 2.6 6.9 4.7 42 86 7556 2.7 2.7 6.8 4.9 56 89
2000:4 7587 2.6 2.7 6.9 4.7 26 87 7605 2.6 2.7 6.8 4.9 42 90
2001:1 7635 2.6 2.8 6.9 4.6 10 87 7654 2.6 2.8 6.8 4.9 27 90
2001:2 7683 2.5 2.8 6.9 4.6 -7 87 7703 2.6 2.8 6.8 4.9 12 91
2001:3 7731 2.5 2.9 6.9 4.6 -25 88 7751 2.5 2.9 6.8 4.9 -4 91
2001:4 7779 2.5 2.9 7.0 4.6 -43 88 7799 2.5 2.9 6.9 4.9 -21 91
2002:1 7826 2.5 3.0 7.0 4.6 -62 87 7847 2.5 3.0 6.9 4.8 -38 91
2002:2 7874 2.5 3.0 7.0 4.5 -82 87 7895 2.5 3.0 7.0 4.8 -56 91
2002:3 7923 2.6 3.0 7.1 4.5 -103 87 7946 2.6 3.1 7.0 4.8 -74 92
Growth rate of PUG-IGZ (2002:3-1996:2)(Congress) -0.5
Growth rate of PUG-IGZ (2002:3-1996:2)(Clinton) 0.1
%GDPR is the percentage change in GDPR (the real growth rate)
%GDPD is the percentage change in GDPD (the inflation rate)
The Congressional experiment has a surplus in fiscal 2002 of $72 billion (the average of 43, 62, 82, and 103 for SGP above), which compares to the CBO estimate in Table 5 of the August 1996 report of $5 billion. Although not shown above, for all four federal revenue categories---personal taxes, corporate taxes, indirect business taxes, and social security contributions---the model has somewhat larger values than does the CBO baseline. From 1999 on the growth rate of GDPR, real GDP, is about 2.5 or 2.6 percent, although for most of 1997 and 1998 it is less than 2.0 percent. The Clinton experiment is similar, although because of the lower spending cuts, the growth rate on average is a little higher and the surplus at the end a little lower. The Clinton surplus in fiscal 2002 is $47 billion (the average of 21, 38, 56, and 74 for SGP above). The unemployment rate by the end rises to 7 percent. The economy does not grow rapidly enough over the period to absorb all the growth of the labor force. Inflation is not a problem, and the bill rate falls slightly. If you are interested in these runs, you should copy BASECONG and/or BASECLIN to a dataset of yours and examine the results in detail. We now turn to Dole's plan. Exogenous Values Used for the Dole Plan The Dole budget numbers were obtained from a table in a "FACT SHEET" put out by the Dole campaign: "Restoring the American Dream: Bob Dole's Pro-Growth Plan for America's Families." The tax reductions in the six fiscal years, 1997-2002, are, respectively, 12, 28, 99, 123, 127, and 129 billion dollars, which total 548 billion dollars. The "additional" spending cuts over the six years (in billions of dollars) are 90 for non-defense administrative costs, 32 for the Energy Department, 15 for the Commerce Department, and 46 for a "1% Reduction in Other Spending Programs," which total 183 billion dollars. I have spread this out over the six years in the same proportion as the total for "Additional Actions" was spread out in the table. In addition to the spending cuts, "Additional Actions" includes 34 billion dollars in revenue from a FCC Spectrum Auction, which is not a spending cut. The cuts I used for the six years are, respectively, 11, 23, 29, 32, 38, and 50 billion dollars. I am assuming that everything else in the Dole table is either an endogenous response of the economy, which the model will estimate itself, or values already incorporated in the Congressional budget resolution. The Dole plan as I am using it is thus 548 billion dollars in tax cuts and 183 billion dollars in spending cuts, both off of the path set by the Congressional budget resolution (i.e., the path used for BASECONG). The tax cuts will all be assumed to be personal income tax cuts and to affect D1G. The spending cuts will all be assumed to affect COG. The base dataset from which the cuts are made is BASECONG. If you have read Chapter 5 of the Workbook, you know how to do this. The following chart gives the calculations (Ch. means "change in," and all flow variables are at quarterly rates): Ch.D1G: Ch.COG:
Ch.Taxes Ch.Taxes/YT Ch.(PG*COG) Ch.(PG*COG)/PG
1996:3 0.0 0 0.0 0
1996:4 0.0 0 0.0 0
1997:1 0.0 0 0.0 0
1997:2 -6.0 -.0041 -2.8 -2.54
1997:3 -6.0 -.0041 -2.8 -2.54
1997:4 -6.0 -.0041 -5.7 -5.11
1998:1 -17.3 -.0115 -5.7 -5.11
1998:2 -17.3 -.0115 -5.7 -5.11
1998:3 -17.3 -.0115 -5.7 -5.11
1999:4 -17.3 -.0115 -7.4 -6.50
1999:1 -27.2 -.0172 -7.4 -6.50
1999:2 -27.2 -.0172 -7.4 -6.50
1999:3 -27.2 -.0172 -7.4 -6.50
1999:4 -27.2 -.0172 -8.0 -6.87
2000:1 -31.9 -.0191 -8.0 -6.87
2000:2 -31.9 -.0191 -8.0 -6.87
2000:3 -31.9 -.0191 -8.0 -6.87
2000:4 -31.9 -.0191 -9.5 -7.97
2001:1 -32.4 -.0183 -9.5 -7.97
2001:2 -32.4 -.0183 -9.5 -7.97
2001:3 -32.4 -.0183 -9.5 -7.97
2001:4 -32.4 -.0183 -12.5 -10.22
2002:1 -32.2 -.0173 -12.5 -10.22
2002:2 -32.2 -.0173 -12.5 -10.22
2002:3 -32.2 -.0173 -12.5 -10.22
The Dole plan begins in 1997:2. The tax cuts are on a calendar year basis, and the spending cuts are on a fiscal year basis. Results for the Dole Experiment The results of solving the model for the Dole assumptions are in dataset BASEDOLE with a password of BASEDOLE. Again, this dataset is like BASE. You can copy it to your own dataset and then examine the results in detail. You can also use your dataset to make different assumptions from what I have done and run a new forecast. A few selected values from the experiment are as follows: Dole Experiment
% %
GDPR GDPR GDPD UR RS -SGP SSP
1996:3 6933 2.8 1.7 5.4 5.3 155 101
1996:4 6970 2.2 1.9 5.5 5.5 168 100
1997:1 7005 2.0 1.9 5.6 5.6 164 99
1997:2 7028 1.3 2.0 5.8 5.5 175 97
1997:3 7055 1.5 2.0 6.0 5.5 171 95
1997:4 7074 1.1 2.0 6.2 5.4 159 92
1998:1 7108 1.9 2.1 6.4 5.3 199 91
1998:2 7147 2.2 2.1 6.6 5.3 194 90
1998:3 7188 2.3 2.1 6.7 5.2 187 90
1998:4 7224 2.0 2.2 6.8 5.1 174 89
1999:1 7270 2.6 2.3 6.9 5.1 200 90
1999:2 7321 2.8 2.3 6.9 5.0 189 91
1999:3 7371 2.8 2.4 6.9 5.0 177 92
1999:4 7419 2.6 2.5 6.9 5.0 165 93
2000:1 7469 2.7 2.6 7.0 5.0 166 95
2000:2 7519 2.7 2.6 7.0 4.9 154 96
2000:3 7569 2.7 2.7 7.0 4.9 143 97
2000:4 7613 2.3 2.8 7.0 4.9 127 98
2001:1 7657 2.4 2.8 7.1 4.9 108 98
2001:2 7703 2.4 2.9 7.2 4.9 95 98
2001:3 7749 2.4 2.9 7.2 4.8 81 99
2001:4 7786 2.0 2.9 7.3 4.8 57 98
2002:1 7828 2.2 3.0 7.4 4.7 35 97
2002:2 7874 2.3 3.0 7.5 4.7 19 96
2002:3 7923 2.5 3.0 7.6 4.7 2 96
Growth rate of PUG-IGZ (2002:3-1996:2) -2.5
%GDPR is the percentage change in GDPR (the real growth rate)
%GDPD is the percentage change in GDPD (the inflation rate)
In general in macroeconometric models government spending changes on goods and services have a bigger impact than do tax cuts, since part of the latter is saved. This property shows up clearly in Table 11.1, p. 282, in Fair (1994), where an increase in COG is noticeably more stimulative in the model than is an equivalent decrease in D1G. This property is important to remember in analyzing the Dole plan. While the decrease in taxes in the Dole plan exceeds the spending cuts, the net effect on real GDP is not that large. If you compare the Dole results to the Congressional results, you can see that GDPR is somewhat higher for Dole in the middle of the period, but that by the end there is very little difference. Also, interest rates are slightly higher for Dole in the middle of the period (because of the slightly more robust economy), and this acts as a slight break on the economy later on. The unemployment rate is noticeably higher for Dole (7.6 at the end versus 7.1), but this is all labor force response. A cut in the personal income tax rate has a positive effect on labor force participation (a supply side feature), which with roughly the same employment growth for both Dole and Congress leads to a higher unemployment rate for Dole. Although output growth is similar in the Dole and Congressional (and Clinton) experiments, the Dole experiment has higher deficits than does the Congressional experiment. We saw above that the Congressional experiment has a fairly large federal government surplus by the end of the period. The Dole experiment essentially eliminates this surplus and has roughly a zero deficit at the end of the period. It is interesting to see what the three experiments imply about -AG, the amount of federal government securities outstanding (i.e., roughly the federal government debt). Although not shown above, the value of -AG in 2002:3 is $4493 billion for the Congressional experiment, $4551 for the Clinton experiment, and $4905 for the Dole experiment. Clinton thus adds $58 billion to the debt from the Congressional value, and Dole adds $412 billion. It should be stressed that the model lacks many of the "supply-side" channels that the Dole campaign is stressing. There is a labor force response, but no automatic way in which the extra workers are employed. Also, there are no increases in tax bases from people sheltering less income or cheating less as tax rates are lowered. The investment response in the model may also be less than a supply-side advocate would believe. From the point of view of the Dole campaign, the above results can probably best be thought of as a lower bound. Conclusion For teachers and students reading this note, I hope I have stimulated you to perform other experiments and to examine the current experiments in more detail. Studying the differences across experiments is a good way of learning about the properties of the model (and maybe about the economy?). You should be able to understand the reasons for the differences by studying the Workbook---I hope there are no black boxes. Which political party should like these results? I suppose there is something for everyone. The Dole campaign can point out that Dole's plan balances the budget by the end of 2002, and the Clinton campaign can point out the Dole plan adds $354 billion more to the debt than does the Clinton plan ($412 billion - $58 billion). The macro implications are similar, and there are clearly not enough big differences among them to wage a campaign on. It is important to remember that all three plans are based on dramatic cuts in government spending from what one would consider normal extensions from historical paths. Hidden behind the macro numbers are large distributional changes, across both generations and income classes. Note also that the macro economy is fairly weak in 1997 and 1998 under all three plans. This is partly caused, of course, by the spending cuts. Reference Fair, Ray C., 1994, Testing Macroeconometric Models, Cambridge, MA: Harvard University Press. |