Tải bản đầy đủ - 0 (trang)
Block 4. Wages, prices and costs

Block 4. Wages, prices and costs

Tải bản đầy đủ - 0trang


competitors in that market. The share of local and external markets then

drives the exports from and imports to the home economy.

Description of the data

REMI has two major uses for data. One is a use for time series data, when

available, to estimate some of the coefficients of the equations. The other

need for data is to calibrate the model to the year that will serve as the base

historical year.

Estimating the coefficients

Due to the completeness of the structure in the REMI model, there is only

a relatively small set of key equation coefficients to be estimated. This is

possible because the model is based on the application of mainstream

economic theory, including the recently developed new economic geography.

The key assumptions underlying the model are that firms seek to maximize

profits, households seek to maximize their well-being, and that workers as

well as goods and services are not homogenous even if they are within the

same occupation or sector of the economy. The equation structure of the

model has been derived from these basic assumptions. In our 20+ years of

experience, it has become evident that, due to the quality of regional data, the

complexity of economics, and the information that firms and households

consider, the larger the data set used the better the quality of the estimates.

This is true even after the structure of the model has been designed in such a

way that the number of coefficients and the data requirements are reduced to

a minimum.

For the quantification at this basic level, the similarity of behavior for

actors in different regions is more important to robust estimation than is the

importance of slight differences from one region to another. In the United

States this has led us to develop a database at the 53-industry level over

25+ years for 3 083 counties. This database obviously includes counties that

run the gamut from small rural counties with only a few thousand people to

populated urban counties. In the estimation process, we filter the data to

insure accuracy for the industry in each county that is included in the

industry-specific estimate.

In Europe, the time series data sets are shorter and often at a higher

geographical level but provide us with a data set that makes it possible to

make any necessary adjustments in our US coefficient estimates due to the EU

specific conditions.

In the rest of this section we will discuss all of the necessary equation

coefficients estimates in logical groups.




We start with two key coefficients used in the market share equations.

βi =The distance (measured as travel time) decay parameter in a gravity

model for industry i.

σι =The price elasticity for industry i.

The βi coefficient determines how the home territories’ share of the demand

in the market of each other territory declines as the travel time to each of the

other territories increases. The σι value shows how the home territories’ share

of each market (including its own market) will change in response to a change

in the delivered cost from the home territory to each of the other territories.

This is relative to the average delivered cost change in each territory based on

its purchases from all of the other territories that it is purchasing from. The

change in delivered cost from the home territory can be due to changes in its

production cost or changes in the cost of delivery.

In order to estimate the βi (travel time decay parameter) dynamically, we

need to have time series estimates for each of the following:

1. An approximation of the change in total output by industry for domestically

produced goods and services in each territory for each year for each


2. An approximation of the change in total demand for domestically produced

goods and services in each territory for every industry and year. This is broken

down into:

a) Consumer demand.

b) Investment demand (gross capital change).

c) Government demand.

d) Intermediate demand.

We can approximate the change in each of these concepts if we have

changes in employment by industry, a national (or EU) input-output table, and a

travel-time matrix from each territory to each other territory. We have obtained

such a matrix commissioned by the EU for all NUTS II and NUTS III regions in

the EU areas from Professor Wegener.7 This is accomplished as follows:

1. An approximation of the change in output for domestic use in each territory

for each year for each industry – use national outputs by industry from the

national input-output table and national employment data (or use time

series national output by industry and employment by industry if available),

then apply the appropriate ratio to the local employment series.

2. A change in total demand by industry for each territory for each year.

a) Consumption demand change – based on changes in the wage bill (or

disposable income if available). Converted to industry demand using

consumption vector in the national or EU input-output table.




b) Investment demand change – based on wage bill in construction (or

changes in the rate of change in the Gross Capital Stock) converted to

industry using the change in gross capital stock in the input-output table.

c) Government demand – changes in the total wages (or changes in

population) convert demands from government from the private

economy to industry demand using the appropriate national inputoutput table.

d) Change in intermediate demand for outputs of each industry based on

output estimates in (1) above and the national (or EU) input-output table.

After this data is assembled, then the βi (travel time decay parameters)

can be found using a search algorithm across the βi values to find the βi values.

This minimizes the rate of error for predicting the change in output in each

territory, based on the changes in demand in each of the other territories for

that industry’s goods or services. For example, output for grocery stores

should be closely related (i.e. a high share) to the changes in demand in the

home area and contiguous areas and not related very closely (i.e. a low share)

to the change in demand in distant areas. However, the output in the home

territory of automobiles should be much more related to demand changes in

all territories than to change in the home area demand (i.e. very low decay)

due to travel time. This will yield a relatively equal share in all markets.

An alternative way to estimate the βi values is possible if one or more sub

national input-output tables are available for any year. This method provides

another quantitative estimate for adjusting the βi values in order to best

explain the inflows and outflows from the local to the other regions based on

one year of data. In the US, we examine the implications of each of the βi

values for the average travel time for the good or service in question relative to

its cost. We also look at the proportion of the local area’s demand served by

the local output over a range of areas. We compared the proportions to those

in similar industries. We estimate the travel time of typical customers for the

particular good or service in question. After doing this analysis we use all of

the information to modify some of the βi values.

The estimate of the σι (elasticity of price response) uses the same time

series data as set forth in the time-series approach for the βi estimates set

forth above. In this case, we use an algorithm to search over values of σι to find

out what value of σι would improve the fit between output changes and

demand changes based on the changes in shares that it would predict in the

markets subsequent to the relative change in production costs (approximated

by wage rate changes) in all of the areas. In other words, the elasticity of

demand σι (the percentage change in the quantity demanded given the

percentage change in relative delivered price) would be determined





In the US, we pool similar industries into categories so that the estimated

elasticity is identical for all of the industries in the broader category. We also

examine the elasticities for reasonableness. For some of the industries, we

change the filter criterion to filter out territories that do not have a substantial

representation of the industry in question or have erratic time series, possibly

due to data reporting and classification errors. Using these methods, we

obtain statistically significant estimates for all industries that meet our own

test for professional reasonableness.

Looking at the employment and wage data available and the national

input-output tables that are available for EU countries, it appears that we will

not have trouble building models for any of these regions.

We typically satisfy the data requirements of the EU models in the

following four ways:

1. Our joint venture partners are provided with a list of our requirements and

are requested to supply as much of the data as they can.

2. We go to the NewCronos database of Eurostat for data that is not supplied

by our partners.

3. We extract data from the Internet and make our own inquiries from any

other sources available. Even with these three sources, however, we often

have missing data points and need to fill in these gaps using estimation


4. In cases where data for previous years are available, fitting a curve to the

available data and extending that curve to the year of interest estimates the

data for the history year of interest. When the data supplied is for a higher

regional level than required, the available data is spread out using

correlated variables that reflect regional variations. A similar spreading is

done when the data available is at an aggregation of the required industries.

In such estimation procedures, we normally have control totals at the

national geographical level, at a sub-national level, and at some aggregation

of industries.

The REMI Policy Insight Model is set up to use very detailed demographic

data, which we have easy access to in the United States. For many European

countries, the data is not available at the level of detail that we need.

Therefore, there are instances when statistical methods are needed to fill in

the missing pieces of data, and we have many means at our disposal to do this.

Usually, population by single age cohort is easily accessible, at least at the

national level. If we have this, then we will use this data to spread out data at

the regional level, if needed, by assuming at the regional level that each age

group within five-year cohort will be in the same proportion to each other, as

they are at the national level.




If we do not obtain local labor force, or don’t have participation rate

forecasts, we will use the participation rates predicted by Eurostat. Since

Eurostat does not have the number of cohorts that we want, we must spread

this data. What we have done in the past is to calculate participation rates

from Eurostat’s population and Eurostat’s labor force. Then we use the same

participation rates for each of the individual cohorts for ages 16 to 19 years old

and likewise for each single age cohort over 75 years old. We will then use the

5-year age and gender groupings for the remaining cohorts.

In the future, we plan to use Eurostat’s population and spread it using the

population data we already have. We will then use Eurostat’s labor force and

spread it, using actual labor force data from a nearby country or the US labor

force proportions, to spread 5 year age cohorts out by single age cohorts. We

can then calculate the participation rates using these values. Next, we will use

these Eurostat-calculated participation rates with the client-provided

population data to calculate a labor force for the forecast. Eurostat only

provides population and labor force forecasts for every five years, so we use a

linear method to fill the participation rates for the intermediate years.

For migration data, we will use national data to spread the age cohorts at

a regional level, or we can use US or another country’s data, spread both to the

national and the regional level if necessary.

In the absence of natality rate forecasts, we assume that natality rates

remain constant over the forecast time period.

For employment by occupation data, we fill the percentages for one

missing industry by using the values from the most similar industry. When we

have no breakdown for manufacturing industries, we use the same value for

all the manufacturing industries that we are given for the manufacturing

industry as a whole.

Using the data that is available to estimate the βi’s and σι ’s in the country

in question, we will examine these estimates in light of the US estimates and

other EU estimates for consistency and reasonableness. Our criteria for

examining the estimates are as follows: the statistical value of the estimates

(e.g. their t-test values), the similarity of the estimates to those in the US and

other EU countries, the consistency of the results from one industry to

another, and the judgment of our contacts in the country in question. We use

our professional judgment on whether to use the new estimates directly, to

use other pooling combinations, or to find other ways to arrive at the best

possible estimates. If no estimates for the particular county are possible due

to a lack of data, we will use other EU estimates combined with US estimates

for β i and σι to make estimates for the country in question. Given the

dominance of many industries by international firms, there is a high

probability that EU elasticities of purchase responses would be fairly




consistent with their responses in the US; therefore, it may be reasonable to

use the US estimates in some cases. Given our techniques of estimation, a

time series estimate can be made with as little as two or three years of data.

Thus, we will have a way to test our estimates in all of the countries under

study to assure that our results are consistent with other countries.

The labor productivity equation uses labor cost σι ’s (estimates of labor

heterogeneity), which are then used in the labor access index. These estimates

were made based on the amount and cost of cross commuting of people in

1300 Traffic Analysis Zones (TAZ’s) around Chicago. The data included the

occupation and industry as well as the place of residence and the place of

work of all those in the sample. We also had travel time and commuting cost

estimates from every TAZ to every other TAZ. The estimates are quite robust

and, since they represent industry hiring decisions, would be expected to be

similar in any economy where there was a benefit in matching workers to

their jobs for the benefit of both the firms and the workers.8

The economic migration speed of response is a key equation to estimate

the European data. In order to estimate this response, we need time series on

the real wage rates, preferably using a deflator that includes housing prices,

the relative employment to labor force ratio, and the number of net internal

migrants each year from region to region. These series, with the possible

exception of the price index using housing prices, are needed to test for an

expected difference in speeds of adjustment for the US and EU. In the absence

of time series data, we could establish whether the speed of adjustment

response should be modified from that in the US or other European countries

by observing the migration flow data that is available.

The labor force equation has two coefficient sets. The first set is related to

the labor force participation responses. This reflects changes in the regions’

current employment relative both to a synthetic labor force using the national

participation rates by cohorts and to the current population by age/gender

cohorts. The other set is related to the relative real wage rate. In the absence of

changes, the participation rate will follow that predicted in the nation for

baseline. The coefficients are by 20 age cohorts for 2 genders. They could be

collectively calibrated to each country if male and female participation rates are

available for a number of years for the regions modeled. There may also be

European studies of participation rate changes caused by real wage and

unemployment rate changes. If these are available, they could be incorporated.

When unemployment rates are high, the general response of decreased

participation rates captures (and the time of adjustment to these changes

reflects) the long-term effects caused by continual high unemployment in

some regions. In the aggregate, we would want to update this equation based

on regional participation rate changes in Europe (or in the country in question)




based on changes in unemployment conditions and the real wage rate. If

possible, we may also test the effects of changes in the real wage relative to

subsidies available for those who are not working.

The housing (or land price) equation will be estimated with data from the

area in question. The change in the price depends on changes in relative real

disposable income and relative population density. The more difficult series to

find is a housing price estimate. This often has to be estimated using a series

compiled by the real estate industry.

Finally, the wage equation depends on the relative moving average of

employment divided by the moving average of the labor force and the industryweighted current occupational demand over its moving average. The data for

this equation requires the industry and employment data mentioned in the first

part of this section in addition to an industry/occupation staffing ratio matrix at

the national level. Initial estimates show somewhat lower wage responses in

the EU than in the US The wage response would need to be re-estimated for the

countries in question, if possible.

Summary for estimates of coefficients

From available evidence, it appears that adequate data exists to estimate the

necessary coefficients for building models for Europe, North America, and most

other market countries. When particular coefficients cannot be estimated for a

certain country, we will have quantitative evidence from other similar countries

that could be used to modify US coefficients as required. We do see the need for

further work to ferret out data that already exists but is not easy to obtain.

Model calibration

While REMI has reduced the number of coefficients that need to be

estimated to a bare minimum, the calibration to a designated last history year

is still an additional task. It involves using all of the information that we have

mentioned above but only for one year. It also involves making a national

forecast for the country or monetary union in which the region is located. We

can do this forecast ourselves, but are be willing to align it with an official

forecast if necessary.

In making our national forecast, we first develop a national labor force

forecast by applying participation rates to cohorts that are consistent with the

Eurostat’s projected participation rates. The combination of our projected

productivity growth by industry and the size of our labor force sets a limit of

output to maintain a fixed employment/labor force ratio for the baseline

national forecast. An upward trend in the employment/labor force forecast

could be built in if it were desirable to assume that current unemployment

rates exceed their likely long-term average rates.




We have produced models of European countries with 24, 26 and

30 industries. In each case, we go to the smallest size region possible (usually

NUTS II or NUTS III) to prepare a database that encompasses the entire

country. This allows us to properly estimate the trade flows at that

geographical level. Then we aggregate these flows for a limited number of

equations and calculate the value of a concept, such as effective distance

(travel time), in such a way that internal flows and the flows from one major

region to another will be consistent with the flows among the smaller regions.

The demographic information necessary for forecast includes detail that

we require for only the region being modeled. This is due to our need to have a

structured demographic model with all the standard demographic processes.

Cost of model construction/use

As implied from the discussion above, REMI Policy Insight® is different

and custom-built for each area we build a model for. Currently, our European

models range is from 24-30 industrial sectors. These include a model of the

Grampian region in the UK, a two-area model of France and the rest of the EU,

a three-area model of the Netherlands, a two-area model of Rhine-Westphalia,

and a single-area model in Tuscany. The latter three models involved ECORYS,

RWI (Rhine-Westphalia Institute for Economic Research, Essen), and IRPET

(Istituto Regionale Per La Programmazione Economica Toscana), respectively.

REMI has recently been chosen by the European Commission to develop our

model methodology for assessing the regional impact of structural funds. We

will be building a model for Southern Italy and a model of four contiguous

regions in Spain as examples for the EU Commission.

REMI Policy Insight is a licensed product customized to the region or

regions for which it is intended. The price is 46 000 euros for a single-area model

of any size region. The price for a two-area model is 53 000 euros. Each

incremental region up to ten regions is 7 000 euros to yield a price of

112 000 euros for a ten-area model. For each of the next 10 areas the increment

is 3 500 euros. After the first year, a 30% maintenance cost provides a new model

each year and continued unlimited telephone and e-mail help in using the

model and interpreting the results.

The use of program evaluations or ex ante projected direct effects

as inputs for the REMI model

REMI Policy Insight® is a macroeconomic forecasting and policy analysis

instrument. It includes a baseline forecast and provides policy variables such

as employment, productivity, taxes, and production costs that can be changed

by the user. Analysts may obtain the program evaluation policy variable inputs

from the outputs of limited program evaluations. They may obtain similar




information from proposed direct interventions for projects under


For example, the direct outputs from traditional cost-benefit analysis of

transportation investments provide the policy variable inputs needed for

transportation studies. These direct variables include such factors as

construction expenditures, operation expenditures, and travel time savings

for businesses and consumers. These variables can be used as inputs into the

REMI model to show the total macroeconomic effects of transportation

infrastructure improvements.

The REMI model has also been linked to specialized microeconomic

models. The Energy 2020 model,9 which provides a high level of detailed

energy-related information, allows users to simulate the effects of different

types of electric generation plants and other energy policies. The outputs of

Energy 2020, including utility construction spending as well as commercial,

industrial, and consumer fuel price changes, are then used as inputs in the

REMI model. The REMI model in turn supplies the predicted change in outputs

that are required by the Energy 2020 model.

Overview of policy analysis areas

The REMI model is a comprehensive forecasting and policy analysis

model. It includes thousands of economic and demographic policy variables

as well as a complete description of the regional economy. Analysts are

therefore able to use the model to evaluate a broad range of policy options.

T h e s e i n c l u d e e c o n o m i c d eve l o p m e n t p ro g ra m s a n d i n c e n t ive s ,

transportation investments, environmental and energy regulations, and other

policies that have an effect on the economy. This section describes

applications in several important areas, with a few illustrative examples that

draw on the thousands of policy analyses that have been conducted using the

REMI model.

Economic development

A broad range of government transfer programs, infrastructure

investments, and business incentive programs are designed for the purpose of

advancing the economic development of local and regional areas. Economic

development as a governmental objective often targets the attraction or

retention of specific firms. Also, infrastructure improvements such as an

airport expansion often have an important economic development


The REMI model has been used for a broad range of economic

development projects. These include issues ranging from the effects of the

horse racing and breeding industry in Minnesota 10 to the impact of a




convention center and hotel in Kansas City11 to the economic impact of casino

gaming proposals in various states.12 Analysts have evaluated the economic

effects of automobile assembly plant locations in Kentucky,13 Michigan,14

Illinois, North Carolina, and Texas. Another theme is the economic effects of

sports stadiums; studies include a football complex in Hartford, Connecticut

and a new baseball park in Boston, Massachusetts.15

Policy makers often evaluate the economic effects of potential losses of

key industries or employers. For example, REMI users have evaluated the

regional economic effects of military base closures,16 declines in tobacco

sales,17 lost coal sales,18 and plant closures.19 Researchers evaluated the

economic effects of land use and growth controls in California. REMI users

have also evaluated economic losses that would occur due to actual or

potential natural disasters such as floods and hurricanes.20

To evaluate a new firm, the analyst typically needs to consider the

construction phase, operational requirements, and government tax or

spending incentives that may have been required to attract the firm. The

construction of a facility occurs for a relatively brief time period, and usually

generates many temporary jobs. Analysts can enter construction in the REMI

model using either a general construction policy variable or policy variables

relating to specific types of construction such as new industrial buildings, new

office buildings, or new commercial buildings.

The user enters the direct effects of the operations phase of a new firm by

using either employment or output policy variables. In using the employment

policy variable, the analyst enters the total number of workers that are directly

employed by the new firm. The employment policy variable assumes a given

level of labor productivity, thus the model adds the direct output associated

with these employees. Similarly, when the user enters output values, the

model calculates the additional direct employment that is needed to produce

this output. The user enters the direct employment or output values before

running the model to show the total economic effects of the new firm.

As part of a targeted economic development policy, government agencies

often offer specific infrastructure investments or tax incentives, or both, as a

means of attracting businesses. For example, a state may provide a new exit ramp

off of an interstate highway as part of a package to attract a manufacturing

facility. More commonly, states and cities may offer corporate profits, property,

and other type of tax breaks in order to bring a facility to a state.

To appropriately capture the economic effects of firm attraction, the

analyst needs to explicitly consider the effects of government investments

and/or tax incentives. For a government infrastructure investment, the

analyst needs to add the infrastructure spending, and elsewhere reduce

government spending or increase taxes to pay for the infrastructure




investment. For a tax incentive, the analyst needs to balance the government

budget by increasing taxes or reducing government spending elsewhere in

order to offset the tax cut.

In some cases, economic development officials may claim that a business

tax incentive that takes the form of a tax exemption does not necessarily

imply that taxes must go up or government spending be reduced elsewhere.

These officials argue that, since a tax exemption involves no government

transfer, it does not require any further government tax or spending changes.

For most tax exemptions, this argument is fallacious. In general, the provision

of government services is closely linked to economic activity. Even businesses

that have few government services requirements often employ workers

whom, as residents, require a high level of educational, fire and safety, and

other public services. Thus, an increase in economic activity requires an

increase in taxation in order to support the need for a higher level of

government services. Without further information, the analyst should assume

that the additional revenues generated by the firm are needed, at least in part,

to pay for additional government service requirements.

REMI model users can implement government infrastructure changes

using policy variables for detailed investments in highways, water and air

facilities, and other detailed expenditure categories. Users can represent

government tax changes with policy variables for personal income taxes,

property taxes, sales taxes (for 13 consumption commodities), and a number

of business taxes.

Some economic development initiatives have special aspects that can be

evaluated using the REMI model. A seasonal or one-time major event such as

a festival or sporting event is often pursued as a means of creating jobs and

economic activity for a city or region. The University of Connecticut evaluated

the economic effects of the proposed 1995 Special Olympics for New Haven,

Connecticut.21 Since much of the construction in preparation for the event

and employment during the event is temporary, the greatest economic effects

are transitory. Long-term effects result from the additional facilities that are

available to the region, as well as possible business-location effects caused by

the publicity generated by the event.

Universities and other institutions are often regarded as catalysts for the

economic development of a region, not only due to the immediate

employment and spending of the university, but also through increasing the

productivity of the labor force and by acting as an incubator for new

technologies and enterprises. Nexus Associates evaluated the effects of Tufts

University School of Veterinary Medicine on the Boston and Massachusetts

economy. In particular, the study traced the economic effects of spin-off

businesses created from University research.22



Tài liệu bạn tìm kiếm đã sẵn sàng tải về

Block 4. Wages, prices and costs

Tải bản đầy đủ ngay(0 tr)