If you have any questions or comments about our FAQ, please feel free to contact us. In referring to particular entries, it is helpful if you also provide the index numbers. Please note that we will periodically update these pages with new material.
The primary advantage REMI PI+ has over I-O models is that it is a dynamic model, which means that it allows for year-by-year analysis, while I-O models are static and do not have time series data. In addition, REMI makes use of Computable General Equilibrium (CGE) techniques, econometric estimations using time series panel data, and the New Economic Geography theory, which takes into account agglomeration effects due to the benefits of access to broader labor and commodity markets.
REMI generally needs between 5 and 10 business days to build and ship a model from the date the contract is received. Models involving subcounty regions or regions outside of the United States may require additional preparation time.
PI+ is customized by region and by the number of industry sectors. REMI can design a single-region model that represents a single county, a group of counties (up to and including a state and additional counties), or even multiple states and additional counties. REMI can also design a multi-region model that can comprise counties or groups of counties. National models as well as sub-county models are also available.
Most of the data for PI+ comes from the BEA, BLS, and Census. However, we use several supplementary data sources. Please refer to the Policy Insight Model Documentation for specific data sources. There is generally a 2-year lag between the current year and the last history year of data.
An I-O table, or Input-Output table, measures the goods that a particular industry buys from all of the other industries (its “inputs”), also known as intermediate inputs. The I-O values include the goods purchased by the intermediate suppliers of the industry. The table can be read as inputs from (industry in row) purchased and converted to output by (industry in column). The values are proportions, so that for every $1 of output by the industry represented by column x, a certain number of cents worth of goods was purchased from each industry y, as given by the value in row y of column x. The I-O table calculates the additional output or jobs that would be created by increasing output for a particular industry; thus, it captures the indirect and induced effects of shocking a specific industry or group of industries.
Economic Geography measures the effects of labor and industry agglomeration. Large labor pools allow firms to hire more specialized workers, leading to lower production costs; similarly, larger supply pools allow firms to purchase more specialized intermediate inputs, thereby also lowering production costs. Access to these broader labor and commodity markets make clustered firms more competitive relative to firms in lower-density economies, which enables them to capture market share and increase employment and output. Economic Geography also allows for the estimation of commuting, transportation, and accessibility costs using the concept of “effective distance.” Distance-decay models suggest that cost is an exponentially increasing function of distance; cost increases by an increasing rate as distance increases. PI+ uses these parameters to gauge the economic impacts of improvements in transportation costs, as entered by users with TranSight.
We currently share out nationwide industry output by each county’s industry compensation.
“Effective distance” is the mechanism through which the theory of economic geography enters the decision-making processes of economic agents in PI+. It adjusts the geographic distance between two centers of economic activity, based on the efficiency of multi-modal transportation between them. Hence, improvements in the transportation infrastructure reduce effective distance between two locations and, consequently, increase their interaction, in terms of the flows of labor, intermediate inputs, and end-use commodities. In general, as effective distance increases, the costs that deter economic activity rise through an exponential process called “distance decay.” The rate of change by economic sector of the distance decay curve (known as the distance decay parameter, ß) captures both the increased deterrence and the variable impact on flows by sector.
This decision depends largely on the format in which your input data is available. From the model’s perspective, units are a matter of indifference, since it can easily convert between orders of magnitude and between real and nominal dollars when performing its calculations. As an example, to switch from real dollars in year X to nominal dollars in year Y, PI+ multiplies the dollar amount by the ratio of the Personal Consumption Expenditure (PCE) index for year Y to the PCE index for year X. You can use the yearly PCE indices from your model results to do similar conversions in a spreadsheet, should you wish to report your forecasts in different units than PI+ uses.
Generally, your data will be available as nominal dollars, which you can enter into PI+ directly without converting it into other units. However, PI+ internally transforms nominal dollars into real dollars using the PCE index or national industry deflators and your region’s realtive delivered price, which might differ from the inflation adjustments embodied in your nominal dollars. To prevent this incompatibility between inflation assumptions, you may prefer to convert your data to real dollars in order to back out inflation. Also remember that if you sustain a flat shock in nominal dollars over multiple years, the shock implicitly loses some value over time. Because of this issue, you may prefer to use real dollars when applying a flat multi-year shock.
Using real dollars for your input values has the advantage of comparability with the model results, many of which are reported in real terms. This comparability often enables you to see the shock directly in your results, or to quickly calculate the implicit multiplier effect. You may notice that for some variables you can choose “fixed” dollars, while for others you can choose “chained” dollars. These are both forms of real dollars pegged to a specific year; for an explanation of the difference, please see the discussion in the Policy Insight Model Documentation.
As predicted by the principles of economic geography, policy changes that enhance a region’s employment opportunities, attractiveness, or market accessibility should stimulate in-migration relative to the baseline. But this can manifest in two ways, depending upon baseline patterns: augmentation of an existing in-migration trend, or attenuation of an existing out-migration trend. If baseline migration is negative (i.e., an outflow) and if the modeled policy produces a negative migration change relative to the baseline (i.e., an escalation of the outflow), then the percentage change in migration will be positive. While this is obviously counterintuitive, it simply reflects that the percentage change is calculated as a negative change divided by a negative level, yielding a positive value. When examining migration results, we recommend focusing on “differences” rather than “percentage differences” to avoid any confusion.
A similar anomaly occurs with the Net Residence Adjustment (NRA) concept, which is found in the Personal Income results table. This number represents the net amount of income that flows into the region when you factor in people who live in the region but commute to jobs in other model regions. If the baseline NRA is negative (which is common for regions that are large industry hubs) and a simulation induces further outflows of income, the percentage change in NRA will appear as a positive value. Because this might be misleading, we advise that you focus on “differences” to see the genuine effect on NRA.
In general, policy variable values represent differences relative to the chosen baseline control forecast. Thus, if you believe the shock’s impact relative to the baseline will be sustained throughout the forecast period, you should leave it in place for all years. For example, the policy variable values embodying employment loss from a military base closure, or the increased production cost caused by a tax increase, must be entered into all forecast years subsequent to occurrence of the shock. If the shock will not be sustained (for example, a temporary surcharge on gasoline designed to raise revenue), the corresponding policy variable should revert to zero one the instigator of the shock disappears.
However, there are three exceptions to this general rule:
If using any of these types of policy variables, you should only enter the shock in the year in which it occurs. The model automatically reflects the change in the remainder of the forecast years.
You should only use the employment update tab when you know the employment for all (or most) industries. This situation might arise when you have more recent information than REMI.
You have two options for updating sector employment in the model. One is to input the revised employment levels for the last historical year and at least the following forecast year, if not several subsequent forecast years. In this case, PI+ will calculate the implicit growth rates and apply them to generate figures for the full forecast period. A second option involves calculating the growth rates in a spreadsheet, developing a full forecast off-line, and then inputting the calculated employment levels directly into the tab. The results should be similar using both methods, although the method used in PI+ calculates growth rates from a base year, not from year to year.
In addition to requiring at least two years of new employment levels, the Employment Update also expects that you enter figures for each of the industries in your model (i.e., each row must be filled in). If you do not have new or revised levels for certain industries, copy the levels from the existing baseline control and paste them into the Update for the appropriate sectors. In addition to copying in the full employment forecast, make sure to copy the last historical year as well. This process updates the employment forecasts for sectors for which you possess data, while retaining the REMI forecasts for the remaining sectors.
When you know the employment in a year for only one industry (such as hotels), you should not use the employment update tab. Instead, you should change a policy variable for the hotel industry. This method ensures that you get all of the indirect and induced affects associated with the new employment level in that industry. Change industry employment for hotels by the difference between the REMI value and the data value, and carry the final observation throughout the remainder of the forecast. Run this through once, and see the results. You will end up with too much hotel employment due to feedbacks to the hotel industry itself. Calculate how much extra there is in hotels, and then run again reducing your initial input by this amount. Repeat this process until you are happy with the hotel-industry employment. This way, all of the industries related to hotels will be adjusted properly.
When simulating a production cost change, consumer prices will automatically change, but the shift will be due to a multiplicity of interacting economic drivers set in motion by the cost shock. If you would like to isolate the impact of the production cost shock on consumer prices, you need to refer to three pieces of data:
1) The change in production cost by industry (in percentage terms)
If you simulated the production cost change using a share variable, this is just your input value. If you entered the cost change as a dollar amount, you need to convert it to a percent by finding the total production costs for an industry for a given year. To do this, divide the static cost change in dollars by the total output of the industry.
2) The local market share of a given industry
This number is already calculated and displayed in the model under Domestic Trade Shares in the “More Tables” button (when looking at regional control results). Use the value of an industry’s trade share within its own region (along the diagonal).
3) The industry share of commodities
This is the most laborious part of this exercise. If you look at the National I-O Matrix and scroll all the way to the right, you see the 79 consumption categories for consumers. The vertical column of coefficients shows how much of each industry goes into these goods (and consequently how much a production cost change will affect its price).
Depending on whether you want to evaluate multiple industries for one commodity, one industry for multiple commodities, or both, you may need to multiply the first two factors (prod cost & market share) by the relevant coefficients for commodities. The result will indicate the price impact of the production cost change.
In some cases, you may want to adjust the intermediate inputs for a particular industry to reflect a different demand vector than that represented by the technical coefficients matrix. This might reflect specific knowledge about a company moving into the region, or about the intermediate inputs to a proposed government undertaking such as a transportation-infrastructure investment.
To do so, enter industry sales for the industry at hand. Note that you must use industry sales rather than firm sales, so that no crowding out occurs (see the question on industry vs. firm for further explanation). Next, select Nullify Intermediate Inputs Induced by Industry Sales / Int’l Exports (amount), using the same value that was entered into sales. This will remove the demand for intermediate inputs that would usually be caused by increasing sales or employment for that industry. Note that you can do this for employment as well, but determining the corresponding changes to intermediate demand is more difficult.
To enter the intermediate inputs, use the Industry Demand, Intermediate Demand (amount) variable, selecting all industries. Input the various amounts of demand for each industry.
The best method for ensuring the integrity of the IO table is to make sure that the total inputs add up to the same proportion of output that they originally did. Thus, if you increase the demand for a particular industry, the combined lowering of other industries must equal that increase. Alternatively, if total intermediate demand is changed from the model’s default calculation, then the variable Value Added with No Effect on Sales or Employment (number) must be adjusted with an equal and opposite change.
Alternatively, you may use the BizDev Blueprint, which has a customize tab to walk you through entering the data necessary to redefine an industry.
The sales and demand variables are essentially two sides of the same coin. Obviously, the decision as to which variables to use depends heavily on the economic shock that you are simulating. But as a general rule, you can choose sales vs. demand based on whether you expect the incremental output to be produced exclusively by local industry or by some combination of firms internal and external to the region. If all new output will be generated by that region’s industry, increase Industry Sales or Firm Sales. But if the output will be provided by a combination of internal and external producers, based on the domestic and foreign trade shares, increase Exogenous Final Demand. Using demand variables is preferable when you don’t know the source of the increased output
To model new investment by businesses or households, you must generally use both the Investment Spending and Capital Stock variables in tandem. First, enter the amount as Investment Spending, which stimulates expansion of output by businesses, leading to growth in employment, wages, and other economic indicators. If the spending occurs over several years, enter each year’s portion into the appropriate year; however, once the spending ceases, do not continue to enter positive values into subsequent years.
The amount of residential or non-residential investment spending you entered will automatically be added to Capital Stock. If the value of the new structure you are simulating is more or less than the sum of the amounts you entered as residential or non-residential investment spending, then you should adjust the Residential or Non-Residential Capital Stock policy variables accordingly. If any of the structure investment was entered as a construction policy variable (either sales or employment) then you must add this to Residential or Non-Residential Capital Stock directly as well.
Firm allows displacement due to competition in the local and nearby markets and the US market. In other words, “crowding out” effects may arise. Industry does not account for any “crowding out” effects. Industry is equivalent to increasing exports to rest of world, and therefore does not compete with the nation or with the particular region. Both firm and industry apply to sales or employment shocks.
Firm employment explicitly factors in displacement or augmentation of existing firms. In a sense, when you reduce employees, other firms hire many of them back, which mitigates the employment impact in the Results. Industry employment changes actually reduce jobs by the full amount input, in addition to causing further reductions through the forward and backward linkages in the input-output table. The concept of firm sales is comparable: when you reduce firm sales, other firms in the same industry expand to fill the void, thus mitigating the negative sales impact in the aggregate. Meanwhile, industry sales reductions actually diminish sales by the entire quantity input.
When deciding whether to shock firm employment vs. industry employment, instead of comparing the effects of the two, determine which is the more relevant policy variable for your situation. You should use firm employment if you believe that the job loss/gain will be at least partially offset by rivals or new market entrants. This is commonly the case for service sectors, for example. But you should use industry employment if the job loss/gain probably won’t be offset, e.g. if the demand is more national or international than local. That is commonly the case with manufacturing, for instance.
If you believe that a portion of the increase will be subject to competition but a portion will be impervious to crowding-out effects, you can do a “hybrid” simulation in which a portion of the employment is entered as firm and a portion as industry.
There are two principal reasons for this. First, employment in the REMI model counts the number of jobs, not the number of employed people. Thus, a person in the labor force who holds two jobs is counted twice toward employment. Second, commuters who reside in other regions can bolster a region’s employment numbers. These workers count toward one region’s employment but another region’s labor force. The model captures commuters through the concept of Residence Adjusted Employment.
Suppose we want to model a higher corporate tax on all industries. We generally recommend against using the Corporate Profit Tax Rate policy variable, since (a) the baseline levels may be inaccurate for certain regions and (b) translating a dollar amount of anticipated revenues into a percentage tax-rate adder (known in PI+ as a “share of tax base”) can be complex. Instead, you should directly increase the cost of capital. This accomplishes the same effect, since the corporate tax rate changes only the cost of capital, which then affects the model through the usual interlinkages. If the tax hike targets selected industries, we can increase the cost of capital by a specific amount for each industry, but for an across-the-board increase (as in this example), we must increase capital costs as a share variable.
Since REMI defines the tax base for the corporate income tax as the capital share of value-added, we need to compute this quantity for each industry, and then translate the tax hike into an appropriate shock.
Depending upon the fuel whose price is increasing, you can manipulate the Fuel Cost policy variables for Electricity, Natural Gas, and/or Residual Fuels, with the latter representing all sources of energy other than electricity or gas. You can enter the increase as either an amount (i.e., the additional fuel-related spending attributable to the higher prices) or as a share increase related to the percentage magnitude of the price hike. The model also permits cost increases targeting particular sectors, or the commercial and industrial sectors as a whole; using the latter shares out the change across component industries proportionally to their usage of the fuel in question.
There are two layers of direct impacts on the production costs of firms. First, the increased cost of one fuel induces substitution toward the other two fuel sources that have now become relatively cheaper, through a Cobb-Douglas process. By stipulating a Cobb-Douglas composition of fuel inputs, the model assumes constant returns to scale and a elasticity of substitution of 1 among the three fuel types. Note that the baseline fuel composition by industry derives from the Fuel Weight Data table, which you can access through the Model Details Button.
Once you alter the fuel composition, overall price-weighted fuel costs generally change from the baseline; in our example, they have probably increased. This triggers a second Cobb-Douglas adjustment to the inputs to production utilized by firms in the affected industries. Specifically, firms substitute away from fuel and toward capital and labor, in accordance with the model’s labor intensity equation (shown in the PI+ Model Equations documentation). Ultimately, the rise in fuel costs translates into increased production costs and hampers the industries’ competitive position relative to industries in regions that do not experience the price change. However, the rising costs are partially mitigated by the ability to substitute toward other fuels and other factors of production.
The best approach is to simulate the higher fuel prices as an increase in the Consumer Price for Electricity, Gas or Fuel & Coal.
Model the migrants themselves as International Migration (under the Population and Labor Supply block). If you have some idea about breakdowns by male/female or age, you can use the detailed cohort variables; otherwise, you can use the All Ages/All Groups variable. To factor in the dependents of these workers, you have to multiply the number of workers by some factor related to the expected average family size of those workers. Enter the total number of migrants in the year you expect the migration to occur only – they are incorporated into the population for subsequent years.
Next, to capture their skills, you should use the Occupational Supply variable to increase the number of trained workers for the occupations you expect the immigrants to fill. Unlike with the migration shock, this increase should be entered for every year in your forecast, assuming you expect these immigrant workers to remain in the labor pool. You can use the “National Industry-Occupation Matrix” (under the Model Details button) to see each occupation’s contribution to the REMI industries. You will see productivity increase in these industries, since they have access to a wider array of qualified workers.
But while the immigrants automatically enlarge occupation pools, you might think that they have higher productivity than the native workers in those same occupations, yet work for the same pay. If that is the case, you could also increase the labor productivity for those industries where the immigrants will find employment. This may be tricky to quantify, since even if the foreign workers in a certain occupation are 10% more productive (for example), the number you enter is scaled down for a) the occupation’s share of the industry’s employment (as seen in the matrix), and b) the immigrants’ share in total employment in the industry (since the other workers aren’t any more productive, all else being equal). So setting this productivity shock requires some thought.
Two other issues as well:
Because retail trade essentially purchases all of its intermediate goods from outside the region (unless your region is very large), and does very little regarding the modification of products, and simply presents a front for selling them, any kind of retail trade study should use the mark-up portion of sales only, no total sales, as the input into firm sales, retail trade in the model. In other words, you should not include the cost of the goods being sold in your policy variable inputs. Otherwise you run the risk of grossly overestimating the benefits of new retail business.
In general, retail trade is not very beneficial to a region for a few reasons:
If some of the goods are manufactured in the area, then you should enter the corresponding value of the goods as an increase in sales for that industry. Example: A new retailer is selling jeans, with estimated sales of $50 million. However, the retailer has a 50% markup, so you should enter only $25 million in the rest of retail industry. However, if the jeans are also produced locally (within the region), then you should allocate the other $25 million to the apparel manufacturing industry.
There are two policy variables that you can use to implement the tax credit, depending on the situation. Under the Compensation, Prices and Costs block, there is an Investment Tax Credit share variable. You can use this variable to implement a percentage-point reduction in the investment tax for all businesses; for example, you would enter a 10% credit as (+) 10 for all years in which the credit will apply. This change lowers the cost of capital and stimulates investment in the model.
If you want to implement a targeted tax credit to a particular company or industry, use the Capital Cost (amount) policy variables. Select the industry or industries you’re targeting (or which include the targeted company), and enter the amount by which your forecasts suggest the credit will reduce their cost (the 10% times expected investment).
You must then offset this credit by either reducing government spending or increasing another tax rate so that the government ledger is balanced. You can do this with suitable policy variables, such as reduced Government Spending if you don’t know which category of government spending will be lowered.
The above procedure yields the economic consequences of the tax credit policy. If you wish to proceed to investigate the fiscal effects (assuming your fiscal revenue and expenditures are reasonably well-calibrated to your own data), enter the Fiscal Calibration policy variable section and adjust both a state-revenue and a state-expenditure variable to reflect the impacts. For example, you can lower Corporate Income Tax relative to the baseline by an amount equaling the reduced investment tax, then lower one or more expenditure categories from which you expect the money to come.
For the sake of simplicity, let’s examine how to model a closure. While simulating the entry of a new firm into the model region is generally analogous, modeling new firms can be tougher since you may not have accurate projections of the firm’s output or employment.
You can elect to reduce either employment or sales for the firm’s industry by the amount which the firm represents. However, be sure that you don’t reduce both or you will double-count the firm closure’s direct effects, as lowering one automatically lowers the other in accordance with industry productivity. Since employment information tends to be more available and reliable than sales data (which is often suppressed or inflated), the best approach is to reduce employment by the number of jobs in the firm. Also, be sure not to remove more employment or output than exist in the baseline, or your results will be unreliable.
The only question is whether to use “firm employment” or “industry employment” as your policy variable. If the company is the only firm of its industry type in the model region, using “industry employment” is preferable. But if there are other firms of its type in the region, you should use “firm employment” since other local firms can pick up a portion of the slack. Please see the FAQ about industry vs. firm for more discussion of this issue.
Inputting a negative shock into an employment policy variable reduces a certain amount of output from that industry, which then translates into two effects that the model captures automatically. First, the shock reduces demand for the industry’s intermediate inputs, with the sectoral breakdown of these reductions based on the industry’s column in the input-output (IO) matrix. Second, industries that used the firm’s output as intermediate inputs satisfy their demand from other sources (other regions or imports). That includes the “diagonal” of the IO, where the firms demanding these inputs are in the same industry as the firm being closed. Again, this effect is automatic and endogenous, so if you then change firm sales or imports, you double-count the effect.
However, this example assumes that the closing firm is a “typical” firm within its industry. If you have information suggesting that its IO parameters would be different than the average firm in the industry, you could apply adjustments to intermediate demand based on the composition of the firm’s actual intermediate inputs. If these adjustments cause a net change in the aggregate input demand, you must balance the input vector by inputting an equal and opposite change into the policy variable “value-added with no effect on sales or employment”. But unless you have information describing how the closing firm’s input composition is different from the average, you could probably ignore this step.
You can perform similar adjustments to the compensation if the company’s average wage exceeds its industry’s average, as reflected in the baseline control. To check this, find the Compensation Rate for the relevant industry in the results of your baseline, and compare it with the average rate that prevailed in the closing firm. Suppose the firm’s average compensation was $1,000 higher than the industry average. When the model takes out employment, it does not remove enough compensation from the affected region. You must input a negative shock to the Compensation (amount) variable, equaling the disparity in the per-person rate ($1,000 in our example) multiplied by the total employees in the firm. This adjustment enables you to fully capture the true income impacts of the firm closure.
The BizDev Blueprint allows you to customize industry data such as intermediate inputs, productivity and compensation, in order to more closely match a specific firm being modeled.
Here are several factors you need to consider when modeling the effects of the closure of a military base.
As always, the selection of appropriate policy variables can depend upon the details of your specific scenario. However, the list below provides guidance on how to simulate basic tax policy changes. Most of the discussion below assumes that you know the dollar impact of the change in tax policy, and thus plan to use Amount policy variables. If you wish to use a share variable, you may need to calculate the percentage price or cost change with reference to an appropriate denominator – for example, total consumer spending on the relevant consumption category, or total output of an industry. Remember that you may wish to match your tax change with an equivalent change in government spending in order to simulate a balanced-budget scenario.
First, view the standard regional control (or the adjusted regional control if you have created a new benchmark) results. Click on the More Tables button of the Economic Tables group and then click on Fiscal to examine the two Fiscal tables (which consist of state revenues and expenditures by category) to determine the existence and magnitude of any discrepancies in the initial year between the model’s data and your own information. The simplest method is to copy the first-year figures into a spreadsheet and calculate percentage differences between model and user data for each revenue or expenditure category from the analyst’s own static projections.
Convert your data to constant dollars of the appropriate base year to be consistent with the units of the fiscal module information. The user may also need to transform fiscal-year data into the calendar-year data used by REMI, which you can accomplish by summing two consecutive fiscal years’ values and dividing the result by two.
Next, create a new regional control. If an adjusted regional control is the desired base, then open the adjusted regional control and edit that. Select fiscal variables from the Fiscal Calibration category for each revenue/expenditure category for which a discrepancy exists. Enter the computed percentage deviations into the corresponding variables for all years of the forecast, by pasting values from the spreadsheet. To enter fiscal calibration changes for input units set to “proportion,” calculate fiscal inputs as user values minus REMI values, divided by REMI value. To enter fiscal calibration changes for input units set to “percent”, calculate fiscal inputs as the user values minus REMI values, divided by REMI value, then multiplied by 100 to generate input changes as percentage. If the user value for a fiscal category exceeds the corresponding REMI value, then input the fiscal variable adjustment as positive; if the user value is less than the REMI value, input the adjustment as negative.
You must apply the adjustment to all years equally in percentage terms, since fiscal forecasts build off their respective base years, which diverged by a known percent. For example, if state general sales tax revenues are low by 3.5% in the first year, enter 3.5% into all years for the State/General Sales Tax policy variable. You may also calculate fiscal adjustment inputs for more than one forecast calendar year, if you have two years of historical data beyond that contained in PI+. In that case, to apply an adjustment through the last forecast year, you must either average the calendar years for which you have calculated input adjustments, or decide which year’s adjustment is the best candidate to apply through the last forecast year. Once this percent adjustment has been implemented, any future movement in the revised fiscal forecast represents indirect effects of endogenous processes in the economic model, such as population shifts.
PI+ essentially consists of two sequential components: an economic engine that produces simulations of economic and demographic effects, and a fiscal module that runs subsequent to the simulation for bookkeeping purposes. To understand the full economic and fiscal impacts of a proposed policy change analysts must use both components. For example, to simulate an increase in the equipment tax, analysts must first capture the economic shock through changes to economic policy variables–specifically, an increased equipment tax rate and increased government spending (if any) due to the incremental tax revenues. The simulation then measures the indirect and induced effects produced by the initial economic shock. Following the economic simulation, the impact on tax revenues is factored into the fiscal module to capture the expected static change in baseline receipts for the relevant tax category. Next, the increased government spending (if any) facilitated by the additional revenue is entered into fiscal-module expenditures, broken down by spending category. Both these stages may require calibration to ensure that fiscal results in PI+ match the user’s projections.
In PI+, “economic” government spending (the policy variable) and “fiscal” government expenditures are defined differently. The government-spending policy variable is designed to capture only those governmental outlays that contribute directly to gross regional product (GRP). By contrast, government spending oriented toward non-productive ends (such as debt service and the redistribution of income) count as budget entries in the fiscal module, but should not be factored into the government-spending economic policy variable. Because of this disparity, quantities entered into the government-spending policy variable and fiscal expenditures may be different.
In developing simulations, the most suitable policy variables for analyzing effects depend on the nature of the policy change under evaluation. To model a tax policy, the user may be able to use a tax-rate policy variable such as the Equipment Tax Rate or Corporate Profit Tax Rate. In cases where these variables are not suitable for the analysis, the user must “disguise” the effect of the tax as an economic concept before incorporating it in the economic model. For example, enter an increased property tax rate as an increase in housing prices based on a static tax amount calculated as some percentage multiplied by the residential capital stock. Or, enter an increased tax on a particular type of capital equipment as either an increased cost of capital or an increased cost of production for the sectors that utilize that equipment. You might enter an increase in a sales tax on a consumer commodity either as a point change in the sales tax or as a static change (after allowing for price elasticity effects on quantity demanded of the commodity) in the tax amount to be collected.
When applying fiscal variables to simulations, insert fiscal variable entries to track tax-related or government-spending related policy variable entries starting from the calendar year in which the fiscal shock occurs in the policy simulation. Then, carry the policy variable and fiscal variable entries for the tax or spending shock in the simulation through the last forecast year, or through the sunset year of the shock, whichever is sooner. In the simulation mode, when using fiscal variables as well as economic policy variables, only one model run is required to properly process the policy variables together with the fiscal variables. However, when creating the simulation, remember to select the baseline containing the adjusted baseline fiscal data.
For the example of an equipment tax hike, there are three direct effects we must incorporate into the model. In the economic modeling, we need to address both the higher equipment tax and the increased governmental spending (if any) that draws from the incremental equipment tax revenues. The third effect involves the post-simulation fiscal balance, which we must restore by adjustments to government tax revenues and expenditures.
First, for the industrial and commercial sector, if the tax applies to the full spectrum of equipment, we can model the tax hike using the Equipment Tax Rate policy variable. Increasing this rate translates into a higher cost of capital and induce substitution away from capital, thereby increasing labor intensities of production.
Second, if the tax is being increased to fund net new spending, we can allocate the increase in government spending to different economic sectors if the funds are earmarked for a specific purpose. If the tax is being increased to cover an operating deficit and thereby merely maintain existing spending, then no spending variables are involved. For example, if incremental tax revenues will be spent on transportation or education, then we can shock the corresponding policy variables. Alternatively, if the government plans on redistributing the income, the analyst could manipulate policy variables involving transfer payments to individuals. In the absence of such specific information, you may simply enter the increase into the “Government Spending-State” policy variable, which allocates those dollars primarily to government payroll and to construction. Remember, however, that the expenditure amount and allocation entered into the fiscal module will likely vary from the policy variable amount, because of their different compositions.
Finally, after running the economic simulation, input tax revenues (based on static projections) and expenditures into the fiscal tracking module. Model the additional equipment tax revenue as receipts under the most suitable tax category (such as “State General Sales Tax Revenue”), and allocate the associated expenditures across categories (such as education and health) based on either general priorities or specifically known earmarkings of the incremental funds. You may need to perform a second round of calibration (as described above) to align the initial-year forecasts of revenues and expenditures with the user’s static projections. Once you’ve entered these percentage adjustments across the full forecast period, remaining differences relative to the static forecast must reflect indirect effects stemming from predicted economic and demographic dynamics.
These categories are defined by the Census Bureau. The composition of these categories is described here.
Before beginning the simulation process, calibrate PI+ baseline fiscal revenues and expenditures in the initial forecast year so that the values are consistent with available actual calendar-year tax receipts and line-item budget data. REMI derives its fiscal segment ratios from Census Bureau data, based on a census of governments conducted at five-year intervals, and an annual survey for the intervening years. REMI averaged census data from two recent fiscal years to create calendar-year ratios, which were applied to the historical data from the model’s last history year.
Because of the data publication lag, the tax activity of legislatures, and the more frequent release of such information within certain political jurisdictions (state, county, etc.), you must recalibrate PI+ to reflect current local data. This fiscal variable calibration process is external to the model; it only affects bookkeeping of fiscal revenues and expenditures, leaving other REMI variables unchanged. For this reason, you must not calibrate the fiscal control until after you have entered and run all other control forecast adjustments. Even if you are concerned only with changes relative to the baseline, these changes may be understated or overstated if the baseline levels are off target.
You can swap RWB files from one computer to another, as long as you swap to the same model on both computers. A workbook should never be transferred to a different version of a model or to a model of a different region.
Similarly, you should not try to import a simulation or adjusted control created in an older version of PI+ into a newer version. If you want to re-use simulations or adjusted controls, you can try to copy the policy variable values (with names and units) from the old simulation and paste them into a fresh simulation in the new version. However, if variable names or numbers have changed between the two versions, the variables may not translate across properly. In that event, you will need to redo the simulation from scratch.
There are several possible explanations for installation difficulties. If none of these applies to your particular problem, please contact REMI.
All simulations are saved in workbook files that have a .RWB file-name extension. It is generally a good idea to maintain several workbooks, organized by project, month, or some other useful separation. If the RWB file gets larger than about 1 GB, it may behave erratically, and become difficult to either save or open. Similar problems may arise even if your workbook is small, if it contains many small simulations.
There are two main ways to keep workbook size within this threshold. First, if you have deleted simulations from your workbook, you may be able to compress down the now-empty space by compacting your workbook.
The best solution is to monitor the size and complexity of your RWB file and start saving your simulations in a new workbook when the old one exceeds 1 GB or becomes excessively cluttered with simulations.
In addition to size-related corruption of the RWB file, there are two other possible causes of opening/saving problems:
Such error messages could indicate any one of a number of potential faults in the model’s performance. To help us to diagnose the problem, please write down the sequence of steps that generated the error, and take a screen shot of the error message itself. To capture the screen shot, press ctrl-PrintScrn while the error message is displayed, open Paint from the Start Menu (located in Programs-Accessories on Windows-based computers), paste the screen shot into the Paint screen (by pressing ctrl-V), and save it as a bitmap (.bmp) file. Email the error sequence and screen-shot bitmap or contact one of our support staff directly.
We recommend that your computer system meet the following specifications. PI+ may run on systems below these minimum specs, but operability is severely impaired. To test whether your memory is sufficiently large, try opening the Standard Regional Control to view the Results; if the Results screen displays successfully, your system should be capable of supporting simulation runs.
1Minimum requirements may be greater depending on model specification
Simulation execution time increases exponentially with the number of regions and industry sectors. Large models (i.e. >25 regions for 23 industries, >12 regions for 70 industries, or >6 regions for 169 industries) will benefit greatly from increased memory capacity and processor speed.
Have you recently changed the date or time on the computer?
When a client changes the date or time on a computer with a keyed model installed, the model becomes “clamped”. It ceases working properly, and the user receives a message upon attempting to open PI+. The model needs to be “unclamped” or unlocked before it will work properly again.
To unclamp a model, follow these steps:
IMPORTANT: Do not close this dialog box until you receive the corresponding “Key” value from REMI.
This may be symptomatic of a minor corruption in your simulation file. Often if you create a new simulation with duplicate variables and values, it will run without a problem. If the error message recurs, please contact us.
The first error happens when the number of iterations required to “solve” the model is different between the baseline control and the simulation. This typically arises when your simulation involves a massive shock, which sometimes requires the simulation to perform additional iterations beyond those required for the baseline control. This means that the forecasts obtained for the baseline and simulation cases are not comparable, since the simulation allowed for additional feedback responses by economic agents. Most likely, this iteration divergence is caused by either incorrect units (for example, dollars entered as units rather than as millions) or by entering a policy variable shock that represents an unrealistically large percentage of the corresponding baseline level. Make sure to check these possible causes first. The second error occurs when the model failed to converge to a solution for at least one forecast year. As with the previous possible cause, you should first check your units and policy variable values to ensure that they are reasonable.
If your inputs are fine, you can increase the “maximum number of iterations” performed by the model for each year, which will allow greater opportunity for convergence. However, don’t increase this number too excessively, as it may significantly slow down the model’s runtime. Your other option is to raise the solution tolerance, which loosens the definition of when convergence has occurred. But as this parameter is less intuitive, we recommend that you try increasing the number of iterations first.
To increase the number of iterations of the baseline control you will need to create a new Regional Control. To do this click on the Regional Control icon under the Add a Forecast group under the home tab. An icon labeled Alternative Model will appear under the Calibration group under the Insert tab. Click on the Alternative Model icon and increase the number of iterations or set a new convergence tolerance level, click OK, and then click on the X at the top right hand corner of the interface just under the resize icons for the interface. This will prompt you to save the baseline control and take you back to the home tab. From there proceed to create your simulation as you did when you got the error, only now select your new control from the baseline options.