Is public expenditure productive? Evidence from the manufacturing sector in U.S. cities, 1880-1920.
Yeoh, Melissa ; Stansel, Dean
This article provides the first examination of the relationship
between public expenditures mad labor productivity that focuses on
municipalities, rather than states or nations. We use data for
1880-1920, a period of rapid industrialization in which there were both
high levels of public infrastructure spending and rapid growth of
productivity. We use a simple Cobb-Douglas production function to model
labor productivity in the manufacturing sector, letting total factor
productivity depend on "productive" public expenditure by city
govermnents--that is, on public spending that may raise the productivity
of labor and encourage human capital accmmulation.
Using a data set of 45 of the largest cities in the United States,
we find no statistically significant relationship between productive
public expenditure and labor productivity in the manufacturing sector
during this period. These findings are robust to three different
econometric approaches. We do, however, find a strongly positive mad
statistically significant relationship between private capital and labor
productivity. Our results are consistent with those of much of the
literature examining this same relationship in states and nations and
they have important implications for contemporary public policy issues.
An Overview
The decline in labor productivity in the United States during the
1970s created a challenging puzzle for economists to solve. Aschauer
(1989) found that public capital had a strongly positive relationship
with productivity in the United States, and argued that the productivity
decline had been caused by a decline in public expenditure on
infrastructure. Munnell (1990) and others found similar results. These
initial Findings were used by politicians and policymakers as evidence
of the need for large increases in government spending on
infrastructure. Some critics identified flaws in the econometric
approach of this early work and, after correcting those flaws, found
either a negative relationship or no statistically significant
relationship between public capital and productivity. Peterson (1994)
found that the marginal rate of return on public capital had declined
substantially since 1950 and was substantially lower than that on
private capital. He suggested that policies to increase private capital
would contribute more to the growth of output than would the increases
in public infrastructure recommended by Aschauer and others. The early
literature on both sides of the debate is summarized in Munnell (1992)
and Gramlich (1994). Few have been able to replicate the large effects
found by Aschauer. Work in this area has slowed, but there remains no
consensus (see, e.g., Kalyvitis and Vella 2011, and Lithgart and Suarez
2011). Moreover, virtually all of the existing literature has focused on
national, regional, or state data and has analyzed contemporary time
periods. While there is a substantial empirical literature investigating
the relationship between local government spending and economic growth,
(1) there appear to be none that examine the relationship with local
labor productivity. (2)
One of Asehauer's (1989: 177) key findings was that
"'a 'core' infrastructure of streets, highways,
airports, mass transit, sewers, water systems, etc. has [the] most
explanatory power for productivity." Since the bulk of that
infrastnleture spending is done by local govermnents, we take a
different approach than previous researchers and focus on local-level
data and do so for a period of rapid industrialization, 1880-1920.
During that time, there was a great deal of public expenditure on the
construction of infrastructure and a rapid increase in the productivity
of labor. (3) If public expenditure is positively associated with
productivity, as some have claimed, then that relationship should be
readily apparent in the data we have chosen.
The period between 1880 and 1920 saw tremendous growth in cities
and wide variations in the labor productivity and public expenditure in
areas across the United States. (4) The United States became a world
leader in manufacturing during this period of rapid industrialization
and much of the industrialization was correlated with city growth. Some
cities recorded rapid population growth rates (for example, Detroit grew
at an average of 73 percent per decade from 1880 to 1920), while other
cities had slower population growth rates (such as Albany's average
of 6 percent per decade).
Since manufacturing generated more than half of the total value of
output in the United States by the late 19th century and was centered in
the largest cities in the nation (Gallman 1960: 9.6), we focus our
attention on whether public expenditure played any significant role in
raising the labor productivity of manufacturing workers specifically.
Labor productivity directly affected the profitability of manufacturing
establishments and thus the overall economic growth of a city. Cities,
in turn, were allocating large quantities of resources toward
"public capital" such as roads, water supply systems,
sanitation, education, and health. Furthermore, local governments were
responsible for the bulk of government activity during this period, so
focusing on city goveruments is most appropriate. (5)
Some argue that public expenditure, particularly in education and
health, increases human capital and thus raises labor productivity in
cities that invested heavily in such areas (Glaeser, Scheinkman, and
Shleifer 1995). However, not every type of public expenditure will raise
labor productivity. Some types of public expenditure, such as spending
on the maintenance of public buildings and the salaries of city
employees in the legislative and judicial branches of government, will
not raise labor productivity in manufacturing. For that reason, we focus
on productive public expenditure.
Economic theory posits that productive public capital--such as
roads, water supply systems, sewers, education, and health--lowers the
cost of doing business and raises the marginal product of other forms of
capital. As a result we should see businesses flourishing in cities that
invested heavily in infrastructure. For example, public expenditure on
roads, bridges, highways, and waterways lowers the cost of
transportation and facilitates the movement of goods and labor
throughout the United States. (6) Public expenditure on education,
health, sanitation, and water supply systems may increase human capital
accumulation by making the labor force (or the future labor force, in
the ease of school children) more literate and healthier. (7)
We can model the growth of a city using the augmented Solow growth
model, assuming that the city is a small economy. This model suggests
that physical and human capital accumulation should go a long way in
explaining the differential income levels of cities. According to Barro
(1997: 2) in his cross-country study of economic growth and convergence,
"The concept of capital in the neoclassical model can be usefully
broadened from physical goods to include human capital in the forms of
education, experience, and health." The effects of physical capital
accumulation (Romer 1986) and human capital accumulation (Lucas 1988) on
economic growth are modeled and documented in many studies, such as
Barro's (1997) cross-country empirical study and Barro and
Sala-i-Martin's (1991) study of income convergence in U.S. states.
(8) Stansel (2005 and 2009) found similar results for the relationship
between human capital and the growth of population and employment in
U.S. metropolitan areas.
Holtz-Ealkin and Schwartz (1995) and Ranch (1994, 1995) provide
formal theoretical models of the relationship between public expenditure
and productivity at the sub-national level that are closely related to
the subject of this article. Those models come to opposite conclusions
about that relationship. Holtz-Eakin and Schwartz's (1995) article
develops a neoclassical growth model explicitly incorporating
infrastructure investments and providing a tractable framework to
empirically analyze the significance of public capital accumulation to
productivity growth. Examining a panel of state data, they find no
statistically significant relationship between public sector capital and
the growth of productivity. Their results suggest that higher
infrastructure outlays were not associated with a significant increase
in productivity growth in U.S. states between 1971 and 1986.
Rauch (1994) develops a formal model to study the effect of
municipal reform in the Progressive Era (from 1902 to 1931) on city
governments' allocation of public expenditure and on city growth,
using the rates of growth in manufacturing employment and value-added
output as measures for city growth. He finds that city governments'
expenditure on roads, sanitation, and the water supply system are
statistically significant in explaining manufacturing employment growth
in both panel and cross-sectional analyses. However, expenditures on
roads and sanitation are not statistically significant in explaining
growth in manufacturing's value-added output in the panel
regression. (9)
Our dependent variable is based on the same data as those used by
Ranch, but we use file level (rather than the growth) of the log of real
dollar value added by manufactures per worker (rather than the total),
that is, we use productivity not overall output growth. Our analysis
differs from Ranch's in four important ways. First, we explicitly
model labor productivity (not output growth) as a function of productive
public expenditure. Second, we include education and health spending in
our public expenditure measure so that it will capture those additional
potential benefits to human capital and thereby productivity. Third, we
examine an earlier time period, 1880-1920 (one that avoids the
potentially contaminating effects of the Great Depression and that
includes the last two decades of the 19th century, which saw large
public investments in basic infrastructure). And, fourth, we do not
examine the impact of municipal reform on cities' growth, which is
the main emphasis of Ranch's analysis. We differ from Holtz-Eakin
and Schwartz (1995) in that we examine public expenditure rather than
the public capital stock, we examine cities rather than states, and we
examine the period 1880-1920 instead of their more recent time period.
We build on the previous literature in this area by providing the
first investigation of the effect of productive public expenditures on
labor productivity in municipalities, which has important implications
for contemporary public policy issues. In recent years, there have been
efforts in the United States and elsewhere to improve economic
conditions by substantially increasing government spending on
infrastructure projects at the state and local level. Proponents of such
efforts have argued that those projects will increase productivity. Our
focus on a period of high public expenditure on physical infrastructure
and rapid industrialization and growth provides an ideal setting for
finding evidence that supports that hypothesis that public expenditure
is productive.
The Theoretical Framework
We follow the lead of previous studies (e.g., Aschauer 1989,
Holtz-Eakin 1994, and Morrison and Schwartz 1992) and specify an
aggregate Cobb-Douglas production function for the manufacturing sector
in city j, for year t, which takes the form:
(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where j indexes the city and t indexes the year. [Y.sub.j,t] is
value added to manufacturing output, [K.sub.j,t] is the value of private
capital, [L.sub.j,t] is the number of workers, and [A.sub.j,t] is the
measure of total factor productivity. We assume that [alpha] + [beta] =
1, implying constant returns to scale. Dividing equation (1) by
[L.sub.j,t] yields the following equation denominated in per worker
units:
(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where [Y.sub.j,t]/[L.sub.j,t] is value added per worker and
[K.sub.j,t]/[L.sub.j,t] is the ratio of private capital to labor. Taking
the natural logarithm of the above equation yields the following:
(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Since local estimates of the value of the public sector capital
stock were not available for this time period, we follow Rauch (1994,
1995) in using productive public expenditure. To investigate the effect
of cities' public expenditure on certain public goods like roads,
water supply systems, sanitation, education, and health, we exclude all
other public expenditure, and we let total factor productivity depend
solely on productive public expenditure as follows:
(4) ln[[A.sub.j,t]] = [A.sub.t] + [gamma][PUBLIC.sub.j,t] +
[c.sub.j],
where At is the time effect common to all cities for a given year,
[c.sub.j] is the city-specific effect, and [PUBLIC.sub.j,t] is the
productive public expenditure per capita in city j in year t. Note that
we are not modeling the effect of public expenditure on private capital
accumulation. The omission of this interaction between public
expenditure and private capital accumulation simplifies the analysis and
allows us to focus on analyzing the levels of public expenditure and
labor productivity for a given level of private capital.
In equation (4) the variable At can be interpreted as the
technology that is available to 'all cities in year t. These time
effects can be consistently estimated with the use of dummy variables
for each year in our sample (YEAR). These year dummy variables are
important because we know that the technology available in the
manufacturing sector changed significantly during the period 1880 to
1920. Therefore, cities chose different levels of public expenditure,
[PUBLIC.sub.j,t], which will 'affect their level of technical
efficiency. (10) We expect to see lower labor productivity in cities
that were slower in installing clean water supply systems and sewers or
cities that invested lower expenditure per capita in the prevention and
treatment of communicable diseases, medical work for school children,
and food regulation and inspection because the residents of such cities
may be less healthy, more prone to diseases, and less productive than
their counterparts in cities that invested early in these public works.
Similar to the time effects, the city-specific effects [c.sub.j]
can be consistently estimated using dummy variables for each city (CITY)
in the sample and omitting one city's dummy variable. These city
dummy variables control for unobservable city-specific factors that do
not change over time but could affect the level of technology in a
specific city. These unobserved city-specific factors could also be
correlated with the explanatory variables and in any given city may
affect labor productivity in an unobservable way. Some examples of
unobservable city effects are agglomeration externalities and knowledge
spillovers for firms located in close proximity to one another within a
city and the level of entrepreneurship in a city. Another unobservable
city-specific effect could be corruption in the city governments and the
presence of patronage politics that affects the level of public
expenditure. Menes (1999) finds that patronage politics in a city
results in higher than optimal provision of public goods and higher
wages paid to city employees.
Finally, substituting equation (4) into equation (3) yields the
following:
(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where [e.sub.j,t] is a random error term. Equation (5) is a
standard two-way fixed effects model with both city and year dummies. We
rename ln[[Y.sub.j,t]/[L.sub.j,t]] as LNVALUE, At as the dummy variable
YEAR, ln[[K.sub.j,t]/[L.sub.j,t]] as LNCAPITAL, and [c.sub.j] as file
dummy variable CITY to yield:
(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
for t = 1880, 1890, 1900, 1910, 1920
for j = 45 cities listed as Albany ... Worcester. (11)
Consequently, equation (6) gives us a theoretical framework in
which to estimate how much, if at all, cities' public expenditure
'affected labor productivity in the manufacturing sector. The
log-normal specification arises because we have assumed an aggregate
Cobb-Douglas production function for the manufacturing sector and we
take the natural logarithm of the production function in order to obtain
a linear equation which we can then estimate using two-stage least
squares (2SLS) regression. The interpretation of the slope coefficient
[gamma] for PUBLIC yields a percentage change in labor productivity
given a dollar change in the level of public expenditure.
One potential problem with this estimation is the endogeneity of
the explanatory variable PUBLIC, that is to say, public expenditure may
be influenced by the sante unobservable factors that influence labor
productivity (i.e., the endogenous variable PUBLIC is correlated with
the model's error term), thus rendering the 2SLS regression
estimates biased and inconsistent (Bound, Jaeger, and Baker 1995: 443).
For example, the demographics of the population or the preferences of
the city governments can endogenously affect public expenditure.
Taxation can also affect both public expenditure and labor productivity
through its impact on private capital accumulation (Rauch 1995: 968-69).
There may be other omitted variables that "also determine labor
productivity and may influence public expenditure.
Another problem is that current expenditures (our independent
variable of interest) depend on per capita income because cities with
higher levels of average income may raise more tax revenues and thus
provide more public goods and this may result in reverse causality
linking the dependent variable (LNVALUE) to the explanatory variable
(PUBLIC).
We deal with this problem of endogeneity in three econometric
specifications: first, we use an instrumental variable (IV) in a 2SLS
estimation; second, we use the initial-year values for the public
expenditures and all other covariates to explain the subsequent 10-year
growth rate of labor productivity (the dependent variable); and third,
we use the lagged public expenditures as an explanatory variable. In our
first method, we use ethnic fragmentation (ETHNIC) as an instrument for
public expenditure because there is existing literature that suggests
that ethnic fragmentation within a city makes it difficult for a city to
agree on public spending due to the heterogeneous preferences of
different ethnic groups over the types of public goods to produce with
tax revenues. Thus, certain public goods such as education, roads, and
sewers supplied by U.S. cities are inversely related to ethnic
fragmentation in those cities (Alesina, Baqir, and Easterly 1999: 1243).
The key finding in Alesina et al. (1999: 1274) is that ethnically
fragmented cities devote lower shares of spending to core public goods
like education and roads.
For our instrumental variable, we use Mesina et al.'s (1999:
1254-55) index of ethnic fragmentation (ETHNIC), which "measures
the probability that two randomly drawn people from a city ... belong to
different ethnic groups." Thus, our measure of ethnic flagmentation
is as follows:
(7) ETHNIC = 1 - [[SIGMA].sub.i] [([Race.sub.i]).sup.2],
where [Race.sub.i] indicates the proportion of the population
listed by the Census as race i and i = {Native White, Foreign White,
African-American, and Other (includes Chinese, Japanese, and American
Indians)}. ETHNIC is a probability that ranges from 0 (if perfect
homogeneity or only a single race lives in a city) to a maximum of 0.75
(if perfectly fragmented into four equally sized racial groups). Our
measure of ethnic fragmentation (ETHNIC) differs slightly from Alesina
et al.'s because we use only four racial groups compared to their
five racial groups, because we grouped Chinese, Japanese, and American
Indians as "Other." The modern Census classification for
"Asian and Pacific Islander" and "Other" (which
largely identifies the Hispanic population in the United States) is
unavailable during the period from 1880 to 1920. The main difference in
our ETHNIC index from Alesina et al.'s is that we separated file
White classification into Native White and Foreign White.
In the 2SLS IV estimation, we use the fitted values (PUBLIC_HAT)
from the first-stage regression of PUBLIC on ETHNIC in file second stage
regression of LNVALUE on PUBLIC_HAT. The idea here is that file
instrumented estimate (PUBLIC_HAT) delivers exogenous variation in the
explanatory variable and allows for a clean identification of the effect
of PUBLIC on LNVALUE. (12) Good instruments should be correlated to the
endogenous variable but should be exogenous or excludable from the
second stage of file 2SLS regression so as not to influence file outcome
directly. The evidence in Alesina et al. (1999) supports our choice of
ethnic fragmentation as an instrumental variable for productive public
expenditure.
The 2SLS regression is specified as follows. In the first stage, we
estimate [PUBLIC_HAT.sub.j,t]:
(8) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
In the second stage, we use the instrumented estimate
[PUBLIC_HAT.sub.j,t] as a regressor:
(9) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
In our estimates of equations (8), (9), (10), and (11), we included
other covariates such as the natural log of city's population, the
natural log of the city's size (in acres), and the natural log of
real wage to control for the effects of these covariates on public
expenditures per capita and on the natural log of value added per
worker. We control for cities' population (LNPOP) because
presumably cities with higher populations would provide more public
goods. We also control for land area in acres (LNLAND) to account for
boundary changes like the annexation of Brooklyn and Allegheny,
respectively, by New York City in 1898 and Pittsburgh in 1907. Ideally
we should be able to control for the level of income per capita by city
(as this will affect the level of public expenditure because cities with
a wealthier tax base may be able to provide more public goods), but
these data do not exist in the time period under study. A good proxy for
personal income per capita by city is the average real manufacturing
wage obtained from the Census of Manufactures by city and deflated by
the national CPI into real 1890 constant dollars. Since wages are
usually assumed to follow a log-normal distribution, we include the log
of cities' real average manufacturing wages (LNWAGE) as a control
in our regression analysis. We also include city-specific (CITY) and
time-specific fixed effects (YEAR).
Our second method to address the endogeneity issue is to use
initial-year values for the public expenditure data (and 'all other
explanatory variables) and subsequent growth (following the initial
year) for the dependent variable. (13) Higher productivity growth from
1880 to 1890, for example, cannot have an impact on the level of public
expenditure in 1880. So we examine in a panel regression the
relationship between public expenditure in the first year of each decade
(1880, 1890, 1900, and 1910) and the growth of labor productivity over
the subsequent decade (1880-90, 1890-1900, 1900-10, and 1910-20),
summarized in equation (10).
(10) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
for t = 1880, 1890, 1900, 1910
for j = 45 cities listed as Albany ... Worcester.
Finally, our third method to address the potentially endogenous
relationship between public expenditure and productivity is to use
lagged values for public expenditure. Productivity in 1890, for example,
cannot have an impact on public expenditure in 1880. Due to limited
availability of data, we must employ a 10-year lag. However, that may be
unrealistically long. Public expenditures in the current year certainly
have a bigger impact on productivity in future years than in the current
year, but the ideal lag may be less than 10 years. Equation (11)
provides the precise specification.
(11) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
for t = 1890, 1900, 1910, 1920
for j = 45 cities listed as Albany ... Worcester.
Data Sources and Construction of Variables
We assemble a panel dataset for 45 of the largest cities in the
United States that spans 40 years by hand-collecting data from the
decennial censuses of 1880 to 1920, with 219 observations for
dries' value added by manufacture, value of private capital in
manufacturing, public expenditure, ethnic fragmentation, average number
of wage earners, city size in acres, and city population. (14) See Table
1 for the summary statistics of the variables used in the regression
analysis and Table 2 for the correlation matrix. We restrict our
attention to the 1880-1920 period for two reasons. This was a period of
rapid industrialization in which there were both high levels of public
infrastructure spending and rapid growth of productivity. Comparable
data were not available for the years before 1880. (15) and
incorporating additional data beyond 1990 would include data
contaminated by the effects of the Great Depression.
Data on the value added by manufacture are published every five
years in the Census of Manufactures and are available by city and by
industry. The value added per worker is file difference in the value of
total output less the cost of raw materials, divided by the average
number of wage earners employed during the year. These wage earners, up
through the working foreman level, are typically production workers,
although there is no way to distinguish them from other nonproduction
wage earners. The figures for value added per worker by city are
deflated by the national Consumer Price Index (CPI) and converted into
the natural log of value added per worker in 1890 constant dollars
(LNVALUE). (16) Similar to other Census data, these value-added data are
subject to some reporting error, but the method of enumeration is
consistent throughout the time period of study. The source for these
value-added data are the Census of Manufactures for the years 1879,
1889, 1899, 1909, and 1919, but these data are also published in the
decennial censuses. Cities' value added by manufacture for the
first three census years (1880, 1890, and 1900) also included "hand
and neighborhood industries" but the latter two years (1910 and
1920) only measure "factory" establishments.
In other studies, such as Mitchener and McLean (1999 and 9,003),
labor productivity is measured by the average manufacturing wage.
However, a problem in using wages as a measure of labor productivity
arises if public expenditure is regarded as an amenity by city
residents, then we can expect to see public expenditure be partly
capitalized into wages (but this may also reflect higher taxes within a
city). The correlation will be negative because people will desire to
live in a city with lots of public goods, thereby bidding the wages
down. This capitalization of public expenditure into wages is the reason
we are using value added per worker in manufacturing as the measure of
labor productivity.
The changes in the level of value added by manufacture or labor
productivity (LNVALUE) are correlated with public expenditure by city
governments. In this article, we use the public capital definition set
out by Corsetti and Roubini (1996: 2) as "government spending
[that] affects the productivity of the final goods sector or the human
capital accumulation sector." Data on city governments'
spending were collected from the U.S. Census volumes on Valuation,
Taxation and Public Indebtedness (1880) and Wealth, Debt and Taxation
(1890) and from volumes of Statistics of Cities and Financial Statistics
of Cities for all cities with populations of 30,000 or more (for 1904)
and 100,000 or more (for 1909 and 1919). Similar to Rauch (1995) our
measure of public expenditure per capita (PUBLIC) is the per capita sum
of city governments' spending on roads, sanitation, and water
supply systems, but we also include education and health spending. We
think it is important to consider these education and health
expenditures, which are not included in Rauch (1995), because they are
components of public expenditure that may increase human capital
accumulation. The expenditures on sewers and water supply systems have
major health implications and thus would also have an effect on human
capital accumulation.
Different types of public expenditure surely affect city
residents' health, education, and labor productivity in different
ways, but we are unable to properly account for these different effects.
Using separate variables for each individual category of spending is
problematic because of inconsistencies in reporting by local
governments. For example, some cities' irregular accounting methods
recorded zero expenditure on water supply systems and schools for some
years (the water works may be contracted out to a private operator and
the school expenditure could be listed under general expenses instead of
under education). Because of those inconsistencies, we take a simple sum
of cities' public expenditure on the five components (roads, water
supply systems, sanitation, education, and health) and divide by the
population of each city to obtain public expenditure per capita
(PUBLIC). We use city population as the denominator because these public
goods are largely nonexeludable and used by the entire population in
each city. PUBLIC is also deflated by the national CPI and converted
into real public expenditure in 1890 constant dollars. We believe the
use of per capita expenditures is superior to the percentage of total
expenditure measure that Ranch uses because it more accurately reflects
the quantity of resources being devoted to public' capital. An even
better measure would be expenditures as a percentage of personal income,
but the persona] income data are not available by city for this time
period.
The Census' of Manufactures ,also reports the aggregate value
of private capital stock, in current dollars, by city and by industry.
These capital figures "show the total amount of capital, both owned
and borrowed, on the last day of the business year reported." (17)
There may be some measurement error in this variable because of the
ambiguity of the census questions and the general difficulties of
accounting for the value of capital goods. Private capital is a
necessary variable to include in our regression analyses because it is
an essential input in the manufacturing sector.
In order to construct the index of ethnic fragmentation (ETHNIC),
we collected data from the U.S. Census volume on Census of the
Population for the number of people within each city who are Native
White, Foreign White, African-American, and Other (includes Chinese,
Japanese, and American Indian) according to the racial classification
used by the U.S. Census. Recall that the index is defined as follows:
ETHNIC = 1 - [[SIGMA].sub.i] [([Race.sub.i]).sup.2],
where [Race.sub.i] indicates the proportion of the population
listed by the Census as race i and i = {Native White, Foreign White,
African-American, and Other (which includes Chinese, Japanese and
American Indians)}. In our data set, the ETHNIC index ranges from a
minimum of 0.15 in Reading, Pennsylvania, in 1880, to a maximum of 0.57
in Memphis for the year 1890. Southern cities--such as New Orleans,
Memphis, Richmond, and Atlanta--show high levels of ethnic fragmentation
because of their larger proportions of African-American residents. These
Southern cities record ethnic fragmentation index numbers between 0.42
and 0.57, whereas cities in the Northeast (except Reading) record ethnic
fragmentation index numbers between 0.28 and 0.50. These Northeast
cities are fragmented between Native Whites and Foreign Whites.
Empirical Results
We use both panel and cross-sectional analyses to investigate the
relationship between cities' productive public expenditure and
labor productivity in the manufacturing sector. Table 3 provides the
2SLS estimates of equations (8) and (9) with robust standard errors. As
panel A indicates for the first stage, in columns (3)-(6) (including all
the regressions where the control variables are included), we see that
ethnic fragmentation has the expected negative sign and is statistically
significant in explaining differences in public expenditure. This
provides evidence that ethnic fragmentation is indeed a good instrument
for public expenditure. It is also economically significant because a
0.10 increase in the ethnic fragmentation index results in $2.15 (in
column (6) specifically) decrease in real public expenditure per capita,
which is about 38 percent of the mean public expenditure per capita.
Panel B illustrates that the instrumented estimate PUBLIC_HAT has no
statistically significant relationship with labor productivity when
private capital and the other control variables are included (columns
(10)-(12)). (18) However, private capital (LNCAPITAL) is highly
significant, both economically and statistically, because a 1 percent
increase in LNCAPITAL is associated with a 40.8 percent increase in
labor productivity. The YEAR dummies are all negative in relation to the
excluded 1920 dummy and are mostly statistically significant at the 1
percent significance level in explaining labor productivity because the
available technology in later years plays a crucial role in increasing
labor productivity.
Our second approach to address the potentially endogenous
relationship between public expenditure and productivity is to regress
initial year values for public expenditure and the other explanatory
variables on subsequent growth of the dependent variable as proposed in
equation (10). Those results are provided in Table 4.
As column (3) indicates, we find no statistically significant
relationship between productive public expenditure and the subsequent
10-year growth of labor productivity. (19) The fact that we are
measuring the growth of productivity rather than the level helps explain
why we find a negative and statistically significant coefficient on
private capital. According to economic theory, cities with higher levels
of private capital will have higher productivity. So, cities with higher
levels of capital in the initial year would be expected to already have
higher levels of productivity. If they have reached the point of
diminishing marginal returns to capital, we would indeed expect to see
lower growth of productivity in the subsequent years, compared to
productivity growth in cities with lower levels of capital.
Finally, our third method for addressing endogeneity is to use
lagged values for the potentially endogenous explanatory variable, so we
regress productivity on public expenditure from 10 years earlier as
proposed in equation (11). None of the other explanatory variables are
lagged. As Table 5 illustrates, once the other covariates are included,
productive public expenditures have no statistically significant
relationship with the productivity of manufacturing labor 10 years
later. This result mirrors our instrumental variables results in Table
3. While we find a small statistically significant positive coefficient
without control variables in column (1), as we add other covariates the
coefficient on LAGGED PUBLIC loses statistical significance and becomes
smaller and changes to a negative sign. As with both of our other sets
of results, the most prominent finding is the strong statistical
significance of private capital.
Our results from these three separate estimation techniques find no
evidence of a statistically significant relationship between public
expenditure mad labor productivity in the manufacturing sector during
the period 1880-1920. These results confirm the theoretical model and
empirical findings of Holtz-Eakin and Schwartz (1995), who document an
insignificant relationship between public expenditure and productivity
growth in states from 1971 to 1986. Rauch (1994) was examining
manufacturing employment growth in cities raffler than productivity, so
our results are not directly comparable. However, our findings are
somewhat at odds with his finding of a positive relationship with some
categories of public expenditures for the period 1902-1931. Finally, our
results are similar to the empirical findings of Glaeser et al. (1995).
Their results indicated that with the exception of sanitation spending,
public expenditures had no relationship with the income and population
growth rates of cities in the period between 1960 and 1990. Our results
also generally confirm their findings for the relationship between urban
growth and racial composition and segregation. More recently, using a
similar approach but examining metro areas instead of cities, Stansel
(2009) also found no significant relationship between local government
spending and economic growth over 1960--90. As discussed previously, the
general consensus in this local growth literature is that government
spending, as a whole, has no significant relationship with growth,
however a positive relationship is sometimes found for specific
categories of spending.
One possible explanation for the consistent finding of an
insignificant relationship is that the potential benefit of local
government spending may be outweighed by the cost of the taxes necessary
to finance that spending. (Unlike the federal government, local
governments generally lack the ability to run a budget deficit.) Taxes
remove money from the private sector, in which the profit motive tends
to ensure its efficient usage, and transfer it to the political sector,
in which electoral motives play a strong role. As a result, when local
government spending and taxes rise, there is a reduction in the
efficiency of the usage of resources. All else equal, areas that utilize
resources more efficiently will tend to have more prosperous economies.
As Stansel (2011) illustrates for metropolitan areas and Poulson and
Kaplan (2008) illustrate for states, higher tax burdens do tend to be
associated with slower economic growth. (20)
Conclusion
This article provides the first examination of the relationship
between public expenditures and labor productivity that focuses on
municipalities, rather than states or nations. Despite examining the
issue in a context very conducive to finding a positive relationship
between public expenditure and productivity (during a period of rapid
expansion of both), this article finds no evidence of such a
relationship.
Once other factors are controlled for, higher levels of productive
public expenditure by city governments have no statistically significant
impact on Tabor productivity in the manufacturing sector for 1880-1920.
These findings are robust to three different econometric approaches.
They are consistent with the findings of much of the other literature in
this area, and they have distinct implications for contemporary public
policy issues.
There have been efforts in many countries in recent years to
dramatically increase public spending as a way to improve economic
conditions. In the United States, for example, federal government
spending increased by more than $1 trillion dollars between fiscal year
2007 (the peak of the previous expansion) and fiscal year 2012, an
increase of 40 percent in just four years. Much of that new spending has
focused on infrastructure projects at the state and local level, with
the argument often being made that those projects will boost
productivity. Our results, and the similar results of others, east doubt
on the ability of that fiscal expansion to achieve its intended effect.
These results may be particularly relevant for rapidly growing
middle-income countries. (21) For local governments in particular,
keeping their tax burden low--especially relative to their
neighbors--may be a more effective strategy for economic revival.
References
Alesina, A.; Baqir, R.; and Easterly, W. (1999) "Public Goods
and Ethnic Divisions." Quarterly Journal of Economics 114 (4):
124,3-84.
Angrist, J. D., and Krueger, A. B. (2001)"Instrumental
Variables and the Search for Identification: From Supply and Demand to
Natural Experiments." Journal of Economic Perspectives 15 (4):
69-85.
Aschauer, D.A. (1989)"Is Public Expenditure Productive?"
Journal of Monetary Economics 23 (2): 177-200.
Barro, B.J. (1991) "Economic Growth in a Cross Section of
Countries." Quarterly Journal of Economics 106 (2): 407-43.
--(1997) Determinants of Economic Growth: A Cross-Country Empirical
Study. Cambridge: MIT Press.
Barro, R.J., and Sala-i-Martin, X. (1991) "Convergence across
States and Regions." Yale Economic Growth Center Discussion Paper
No. 629. New Haven, Conn.: Yale University.
Black, D., and Henderson, V. (1999) "The Theory of Urban
Growth." Journal of Political Economy 107 (2): 252-84.
Boarnet, M. (1998) "Spillovers and the Locational Effects of
Public Infrastructure." Journal of Regional Science 38 (3):
381-400.
Bound, J.; Jaeger, D.A.; and Baker, R. M. (1995) "Problems
with Instrumental Variables Estimation When the Correlation between the
Instruments and the Endogenous Explanatory Variable Is Weak."
Journal of the American Statistical Association 90 (430): 443-50.
Corsetti, G., and Roubini, N. (1996) "Optimal Government
Spending and Taxation in Endogenous Growth Models." NBER Working
Paper No. 5851. Cambridge: National Bureau of Economic Research.
Crihfield, J.B., and Panggabean, M.P.H. (1995) "Growth and
Convergence in U.S. Cities." Journal of Urban Economics 38 (2):
138-65.
Dalenberg, D.R., and Partridge, M.D. (1995) "The Effects of
Taxes, Expenditures, and Public Infrastructure on Metropolitan Area
Employment." Journal of Regional Science 35 (4): 617-40.
De Mello, L.R., Jr. (2002)"Public Finance, Government Spending
and Economic Growth: The Case of Local Governments in Brazil."
Applied Economics 34 (15): 1871-83.
Denaux, Z. S. (2007) "Endogenous Growth, Taxes and Government
Spending: Theory and Evidence." Review of Development Economics 11
(1): 124-38.
Deno, K. T. (1988) "The Effect of Public Capital on U.S.
Manufacturing Activity: 1970 to 1978." Southern Economic Journal 55
(2): 400-11.
Duffy-Deno, K. T., and Eberts, R. W. (1991) "Public
Infrastructure and Regional Economic Development: A Simultaneous
Equations Approach."Journal of Urban Economics 30 (3): 329-43.
Eberts, R. W. (1986) "Estimating the Contribution of Urban
Public Infrastructure to Regional Growth." Federal Reserve Bank of
Cleveland Working Paper No. 8610.
--(1990) "Public Infrastructure and Regional Economic
Development." Federal Reserve Bank of Cleveland Economic Review 26
(1): 1.5-27.
Gallman, R. E. (1960) "Commodity Output, 1839-1899." In
National Bureau of Economic Research, Trends in the American Economy in
the Nineteenth Century, 13-72. Princeton: Princeton University Press.
Glaeser, E. L.; Scheinkman, J. A.; and Shleifer, A. (1995)
"Economic Growth in a Cross-Section of Cities." Journal of
Monetary Economics 36 (1): 117-43.
Glaeser, E. L., and Shapiro, J. (2003) "Urban Growth in the
1990s: Is City Living Back?" Journal of Regional Science 43 (1):
139-65.
Gramlich, E. M. (1994) "'Infrastructure Investment: A
Review Essay." Journal of Economic Literature 32 (3): 1177-96.
Holtz-Eakin, D. (1994) "Public Sector Capital mad the
Productivity Puzzle" Review of Economics and Statistics 76 (1):
12-21.
Holtz-Eakin, D., and Schwartz, A. E. (1995) "Infrastructure in
a Structural Model of Economic Growth." Regional Science and Urban
Economics 25 (2): 131-51.
Jiwattanakulpaisarn, P.; Noland, R. B.; Graham, D. J.; and Polak,
J. w. (2009) "Highway Infrastructure Investment and County
Employment Growth: A Dynamic Panel Regression Analysis." Journal of
Regional Science 49 (2): 263-86.
Kalyvitis, S. and Vella, E. (2011) "Public Capital
Maintenance, Decentralization, and U.S. Productivity Growth."
Public Finance Review 39 (6): 784-809.
Kendrick, j. (1984) "U.S. Economic Policy and Productivity
Growth." Cato Journal 4 (2): 387-400.
Lithgart, J. E., and Martin Suarez, R. M. (2011) "The
Productivity of Public Capital: A Meta-analysis." In W. Jonkhoff
and W. Manshanden (eds.) Infrastructure Productivity Evaluation, 5-32.
New York: Springer.
Lucas, R. E., Jr. (1988) "On the Mechanics of Economic
Development." Journal of Monetary Economics 22 (1): 3-42.
Menes, R. (1999) "The Effect of Patronage Politics on City
Government in American Cities, 1900-1910." NBER Working Paper No.
6975. Cambridge: National Bureau of Economic Research.
Mitehener, K. J., and McLean, I. W. (1999) "U.S. Regional
Growth and Convergence, 1880-1980." Journal of Economic History 59
(4): 104G42.
--(2003) "The Productivity of U.S. States Since 1880."
NBER Working Paper No. 9445. Cambridge: National Bureau of Economic
Research.
Moomaw, R. L., and Williams, M. (1991) "Total Factor
Productivity Growth in Manufacturing: Further Evidence from the
States." Journal of Regional Science 31 (1): 17-34.
Morrison, C. J., and Schwartz, A. E. (1992) "State
Infrastructure and Productive Performance.'" NBER Working
Paper No. 3981. Cambridge: National Bureau of Economic Research.
Munnell, A. H. (1990) "Why Has Productivity Declined?
Productivity and Public Investment." Federal Reserve Bank of
Boston, New England Economic Review (January-February): 3-22.
--(1992) "Infrastructure Investment and Economic Growth."
Journal of Economic Perspectives 6 (4): 189-98.
Peterson, W. (1994) "Overinvestment in Public Sector
Capital." Cato Journal 14 (1): 6,5-73.
Poulson, B. W., and Kaplan, J. G. (2008) "State Income Taxes
and Economic Growth." Cato Journal, 28 (1): 53-71.
Rauch, J. E. (1994) "Bureaucracy, Infrastructure, and Economic
Growth: Theory and Evidence from U.S. Cities during the Progressive
Era." Department of Economics Discussion Paper No. 94-06. San
Diego: University of California, San Diego.
--(1995) "Bureaucracy, Infrastructure, and Economic Growth:
Evidence from U.S. Cities during the Progressive Era." American
Economic Review 85 (4): 968-79.
Reed, W. R. (2008) "The Robust Relationship between Taxes and
U.S. State Income Growth." National Tax Journal 61 (1): 57-80.
Romer, P. M. (1986) "Increasing Returns and Long-run
Growth." Journal of Political Economy 94 (5): 1002-37.
Stansel, D. (2005) "Local Decentralization and Local Economic
Growth: A Cross-Sectional Examination of U.S. Metropolitan Areas."
Journal of Urban Economics' 57 (1): 55-72.
--(2009) "Local Government Investment and Long-Run Economic
Growth." Journal of Social, Political, and Economic Studies 34 (2):
244-59.
--(2011) ""Why Are Some Cities Growing While Others Are
Shrinking?" Cato Journal 31 (2): 285-303. U.S. Department of
Commerce, Bureau of the Census (1880) Valuation, Taxation and Public
Indebtedness. Washington: U.S. Government Printing Office.
--(1890) Wealth, Debt and Taxation. Washington: U.S. Government
Printing Office.
--(1904) Statistics' of Cities Having a Population of over
30,000. Washington: U.S. Government Printing Office.
--(1910, 1921) Financial Statistics of Cities Having a Population
of over 30,000. Washington: U.S. Government Printing Office.
--(1975) Historical Statistics of the United States', Colonial
Times to 1970. Washington: U.S. Government Printing Office.
--(2008) Historical Overview of U.S. Census Bureau Data Collection
Activities about Governments: 1850 to 2005. Washington: U.S. Government
Printing Office.
--(various years) Census of Manufactures. Washington: U.S.
Government Printing Office.
--(various years) Census of the Population. Washington: U.S.
Government Printing Office.
Walton, G. M., and Rockoff, H. (2002) History of the American
Economy. 9th ed. Toronto: Thomson Learning.
(1) See, for example, Glaeser, Scheinkman, and Shleifer (1995),
Holtz-Eakin and Schwartz (1995), Crihfield and Panggabean (1995),
Dalenberg and Partridge (1995), De Mello (2002), Glaeser and Shapiro
(2003), Denaux (2007), and Stansel (2009). The general consensus of this
literature is that government spending, as a whole, has no significant
relationship with growth, however a positive relationship was found for
several specific categories of spending.
(2) Rauch (1994), Eberts (1986), Deno (1988), Eberts (1990),
Duffy-Deno and Ebex'ts (1991), and Boarnet (1998) all take a local
approach and are closely related to the issue we examine, but none
contain results using local labor productivity as their dependent
variable.
(3) As Kendrick (1984: 389) documents, the productivity of labor
over a similar period (1889-1919) was nearly double that of the previous
four decades (185.5-90). Total productivity was more than five times
higher.
(4) In an empirical study on states, Mitchener and McLean (2003:
34-3.5) document "'massive and persistent differences in
productivity levels, and hence living standards" across the 48 U.S.
states (excluding Alaska and Hawaii) from 1880 to 1960.
(5) In 1902, local governments accounted for 5.5 percent of all
government revenue and 59 percent of 'all government outlays,
compared to about 22 percent and 2.5 percent today (Menes 1999).
(6) Moomaw and Williams (1991) found some evidence of a positive
relationship between highway infrastructure and productivity in the 48
contiguous states. However, Jiwattanakulpaisarn et al. (2009) found no
statistically significant relationship between investments in highways
and employment in 100 counties in North Carolina.
(7) Menes (1999: 2) states, "The roads, sewers, schools,
transportation, electricity, gas and water provided by local governments
or by government franchisees were vital to the health, wealth, and
happiness of residents.'"
(8) Black and Henderson (1999) provide a formal model of human
capital accumulation and urban growth.
(9) Rauch (1995: 969) states, "Investment in new
infrastructure is assumed to generate city growth by providing a
complementary input that attracts investment of private capital in
traded goods industries (manufacturing), creating jobs which in turn
attract migrants from a surrounding agricultural hinterland."
(10) For simplicity, following Glaeser et al. (1995) and others, we
ignore the impact of the revenue source required to finance this higher
spending (taxes or bonds). Since higher taxes would tend to reduce
productivity, this implicitly biases upward our coefficients on public
expenditure.
(11) Our sample consists of 45 of the largest U.S. cities: Albany,
Atlanta, Baltimore, Boston, Buffalo, Cambridge, Chicago, Cincinnati,
Cleveland, Columbus, Dayton, Detroit, Fall River, Grand Rapids,
Hartford, Indianapolis, Jersey City, Kansas City, Louisville, Lowell,
Memphis, Milwaukee, Minneapolis, Nashville, New Haven, New Orleans, New
York, Newark, Omaha, Paterson, Philadelphia, Pittsburgh, Providence,
Reading, Richmond, Rochester, San Francisco, Scranton, St. Louis, St.
Paul, Syracuse, Toledo, Trenton, Wilmington, and Worcester.
(12) See Bound, Jaeger, and Baker (1995) and Angrist and Krueger
(2001) for an introduction to IV estimation.
(13) There is precedent for this in the literature. For example,
Barro (1991) and Glaeser et al. (1995) both use 1960 government
expenditure data as an explanatory variable for 1960-90 economic growth.
(14) We are missing observations for Grand Rapids, Memphis, Omaha,
and Trenton in 1880. We also dropped two observations that were
substantial outliers (Wilmington in 1880 and Reading in 1920, which were
not representative of the usual trend of annual public capital spending)
and thus would distort the coefficient estimates. In 1880, Wilmington
spent 581 percent and 176 percent more on roads and education,
respectively, in real terms compared with the average of 1890 to 1920.
Similarly, in 1920 Reading spent 64 percent more on sewers, in real
terms compared with the average between 1880 and 1910. Reading's
spending on education cannot be consistently compared due to accounting
irregularities (zero dollars recorded for 1880 and 1890 and $4,500
current dollars in 1910).
(15) See U.S. Department of Commerce (2008) for a discussion of the
history of the Census Bureau's collection of data on U.S.
governments.
(16) The national CPI--constructed by the Bureau of Labor
Statistics--is available in the Historical Statistics' of the
United States, Colonial Times to 1970 (Series E135-166). We use a
national CPI to make price adjustments because during 1880-19'20,
the commodities and labor markets in the United States were pretty well
integrated due to the completion of die transcontinental railway (Walton
and Rockoff 2002: 363).
(17) The quote is excerpted from the "Explanation of
Terms" from the Census of Manufactures (1920: 1.5-18).
(18) We also estimated equation (5) by cross-sectional regressions
for each year. The cross-sectional regression allows the constant and
slope coefficients to change for each year. The different signs, sizes
of the coefficients, and significance levels make it harder to formulate
an overall interpretation for the entire time period under study, but we
found that public expenditure per capita is not statistically
significant with respect to labor productivity in each of the years
studied. For brevity, those results are not included herein.
(19) Running the regressions as four separate cross-sections,
instead of as a panel, yields similar results. All coefficients on
public expenditure are statistically insignificant as in Table 4. The
one difference is that the coefficient in the regression for 1880-90
productivity growth has a positive sign. For brevity, those results are
not included herein.
(20) Reed (2008) provides an excellent summary of the voluminous
literature in this area, focusing on the states.
(21) We are grateful to an anonymous referee for making this point.
Melissa Yeoh is an Assistant Professor of Economics at Berry
College, and Deem Stansel is an Associate Professor of Economics at
Florida Gull" Coast University. They thank Noel Campbell, Garth
Heutel, Brian McNamatra, Mitch Mitchell, and Bob Mulligan for helpful
comments on a previous version of this article and Amy Weisgarber for
research assistance. Yeoh's work was supported by the Kirk Dombush
Summer Research Grant at Vanderbilt University. She thanks Jeremy Mack,
Martha Bailey, Bill Collins, Tomas Cvrcek, Yanqin Fan, William
Hutchinson, Bob Margo, Mototsugn Shintani, and John Warner for comments
on earlier drafts.
TABLE 1
DESCRIPTIVE STATISTICS
Std.
Variable N Mean Dev. Min Max
Ln Value Added 219 7.05 0.35 6.07 8.01
Public Expenditure 219 5.61 2.55 0.28 15.39
Ethnic Fragmentation 219 0.43 0.07 0.15 0.57
Ln Private Capital 219 7.50 0.48 6.12 8.61
Ln City Population 219 12.13 0.96 10.53 15.54
Ln City Size (acres) 219 9.67 0.88 8.05 12.25
Ln Real Wage 219 6.14 0.21 5.29 6.64
NOTE: Year and city dummies are excluded from this table.
TABLE 2
CORRELATION MATRIX
Ln Ln
Value Public Ethnic Private
Added Expenditure Fragmentation Capital
Ln Value Added 1
Public Expenditure 0.369 1
Ethnic Fragmentation -0.158 0.018 1
Ln Private Capital 0.850 0.389 -0.139 1
Ln City Population 0.524 0.300 0.137 0.422
Ln City Size (acres) 0.297 0.194 0.200 0.226
Ln Real Wage 0.814 0.382 -0.178 0.729
Ln Ln
city city Ln
Population Size (acres) Real Wage
Ln Value Added
Public Expenditure
Ethnic Fragmentation
Ln Private Capital
Ln City Population 1
Ln City Size (acres) 0.762 1
Ln Real Wage 0.453 0.283 1
TABLE 3
TWO-STAGE LEAST SQUARES REGRESSION RESULTS
(1) (2) (3) (4)
First-stage dependent variable: PUBLIC
ETHNIC 0.621 3.333 * -20.56 *** -21.46 ***
(2.344) (1.881) (6.196) (6.148)
PUBLIC HAT
LNCAPITAL 1.669 **
(0.644)
LNPOP
LNLAND
LNWAGE
1880 DUMMY -1.988 *** -0.744 1.137
(0.472) (0.522) (0.915)
1890 DUMMY -1.239 *** 0.0975 1.177 *
(0.361) (0.425) (0.602)
1900 DUMMY 2.122 *** 2.832 *** 3.352 ***
(0.360) (0.342) (0.407)
1910 DUMMY 0.318 0.969 ** 0.997 **
(0.431) (0.446) (0.436)
CONSTANT 5.347 *** 4.292 *** 11.24 *** -1.620
(1.003) (0.784) (1.948) (5.1.59)
CITY DUMMIES? No No Yes Yes
Observations 219 219 219 219
R-squared 0.000 0.298 0.619 0.629
(5) (6) (7) (8)
First-stage Second-stage dependent variable:
dependent LNVALUE
variable:
PUBLIC
-21.50
(6.056)
ETHNIC -21.12 -1.239 ** 0.0110
(6.077) (0.581) (0.0598)
PUBLIC HAT 1.591 **
LNCAPITAL 1.475 (0.668)
(0.684) 1.004
LNPOP 0.993 (0.698)
(0.703) 0.190
LNLAND 0.130 (0.509)
(0.497) -0.940
LNWAGE -1.637
1880 DUMMY 2.070 * 1.860 -0.832 ***
(1.109) (1.203) (0.124)
1890 DUMMY 1.783 ** 1.793 ** -0.454 ***
(0.725) (0.723) (0.0781)
1900 DUMMY 3.759 3.731 *** -0.347 **
(0.490) (0.499) (0.138)
1910 DUMMY 1.203 1.186 *** -0.206 ***
(0.450) (0.455) (0.0507)
CONSTANT -13.30 -8.858 14.00 *** 7.343 ***
(8.418) (11.57) (3.263) (0.338)
CITY DUMMIES? Yes Yes No No
Observations 219 219 219 219
R-squared 0.633 0.634 0.025 0.633
(9) (10) (11) (12)
Second-stage dependent variable: LNVALUE
ETHNIC -0.0092 0.00319 -0.00067 -0.0090
(0.0218) (0.0169) (0.0175) (0.0158)
PUBLIC HAT 0.471 *** 0.449 *** 0.408 ***
LNCAPITAL (0.0652) (0.0646) (0.0698)
0.0763 0.0799
LNPOP (0.0581) (0.0512)
0.06101 0.0345
LNLAND (0.0348) (0.0315)
0.433 *
LNWAGE (0.126)
1880 DUMMY _0.860 *** -0.319 *** -0.242 ** -0.128
(0.0533) (0.0737) (0.0939) (0.0833)
1890 DUMMY -0.471 *** -0.167 *** -0.109 * -0.0991 *
(0.0378) (0.0418) (0.0558) (0.0506)
1900 DUMMY -0.297 *** -0.186 *** -0.135 ** -0.0911
(0.0625) (0.0547) (0.0617) (0.0585)
1910 DUMMY -0.193 *** -0.197 *** -0.172 *** -0.155
(0.0352) (0.0248) (0.0278) (0.0291)
CONSTANT 7.392 *** 3.621 *** 2.337 *** 0.180
(0.117) (0.498) (0.802) (0.813)
CITY DUMMIES? Yes Yes Yes Yes
Observations 219 219 219 219
R-squared 0.879 0.924 0.927 0.933
NOTE: Robust standard errors in parentheses.
*** p<0.01, ** p<0.05, * p<0.1.
TABLE 4
REGRESSION RESULTS FOR TEN-YEAR
GROWTH RATES IN LNVALUE
(1) (2) (3)
Dependent variable: DLNVALUE
PUBLIC -0.00964 * -0.00768 -0.0125
(0.00582) (0.00632) (0.00911)
LNCAPITAL -0.120 *** -0.187 *** -0.265 **
(0.0445) (0.0658) (0.125)
LNPOP 0.0404 ** 0.03880 * 0.218 **
(0.0198) (0.0190) (0.0992)
LNLAND -0.0109 -0.0204 -0.0956
(0.0214) (0.0232) (0.0726)
LNWAGE -0.192 ** -0.0248 -0.333
(0.0964) (0.101) (0.247)
1890 DUMMY -0.150 *** -0.0988
(0.0388) (0.0915)
1900 DUMMY -0.0934 -0.0327
(0.0586) (0.120)
1910 DUMMY 0.0194 0.0754
(0.0772) (0.164)
CONSTANT 1.948 *** 1.571 *** 2.565
(0.487) (0.596) (2.039)
CITY DUMMIES? No No Yes
Observations 174 174 174
R-squared 0.245 0.356 0.515
NOTE: Robust standard errors in parentheses.
*** p<0.01, ** p<0.05, * p<0.1.
TABLE 5
REGRESSION RESULTS WITH LAGGED PUBLIC EXPENDITURES
(1) (2) (3) (4)
Dependent variable: LN VALUE
LAGGED PUBLIC 0.0317 *** 0.003 -0.00158 -0.0017
-0.00578 -0.00561 -0.0052 -0.00519
LNCAPITAL 0.585 *** 0.526 *** 0.524 ***
-0.0519 -0.0556 -0.0564
LNPOP 0.0787 *** 0.102 ***
-0.0149 -0.0286
LNLAND -0.0295
-0.0261
LNWAGE
1890 DUMMY
1900 DUMMY
1910 DUMMY
CONSTANT 6.984 *** 2.655 *** 2.172 *** 2.180 ***
-0.0395 -0.38 -0.376 -0.38
CITY DUMMIES? No No No No
Observations 174 174 174 174
R-squared 0.104 0.569 0.627 0.63
(5) (6) (7)
Dependent variable: LN VALUE
LAGGED PUBLIC -0.00533 -0.000637 -0.00429
-0.00479 -0.00447 -0.00434
LNCAPITAL 0.420 *** 0.394 *** 0.448 ***
-0.0596 -0.0882 -0.0882
LNPOP 0.0748 *** 0.0645 *** 0.217
-0.0223 -0.0218 -0.0743
LNLAND -0.0269 -0.022 -0.000298
-0.0235 -0.0226 -0.034
LNWAGE 0.645 *** 0.584 *** 0.221
-0.131 -0.126 -0.16
1890 DUMMY -0.1101 ** -0.00901
-0.0537 -0.0527
1900 DUMMY -0.122 *** -0.0736 **
-0.041 -0.0356
1910 DUMMY 0 -0.136 ***
-0.0325 -0.0313
CONSTANT -0.69 0.0337 -0.114
-0.662 -0.734 -0.809
CITY DUMMIES? No No Yes
Observations 174 174 174
R-squared 0.704 0.749 0.911
NOTE: Robust standard errors in parentheses.
*** p<0.01, ** p<0.05, * p<0.1.
