Plasmodium Vivax

The parasite Plasmodium, which is the cause of malaria, was first discovered in 1880. Malaria is transmitted when the parasite “… is injected into the human blood stream by the bite of an infected mosquito. Shortly thereafter, the parasite enters a red blood cell and begins to grow and multiply until from 16 to 24 new parasites are formed. The red cell then bursts, freeing the parasites which soon enter other red cells to undergo similar development.”Footnote ²

Infection can be devastating. Anemia results from the parasite's destruction of red blood cells. In mild or uncomplicated cases of malaria, cold chills are followed by fever and nausea, and eventually the breaking of the fever. In more complicated cases, blood may appear in one's urine. Pulmonary edema, reduction in blood platelets, and more severe complications may lead to death. Plasmodim Vivax, the variety most prevalent in the United States may remain dormant in the blood for several years, causing complications well after the initial transmission.

Breeding Grounds and Eradication Efforts

The Anopheles mosquito is responsible for the transmission of plasmodium vivax in the United States. These mosquitoes lay their eggs on sitting fresh water, so any idle pool (inlets, sink ponds, and swamps) is a potential breeding ground. When reservoirs are built behind dams sitting water may increase or decrease. Previous breeding grounds adjacent to the river become flooded and leave larvae susceptible to natural prey. On the other hand, increases in shoreline often create larger breeding grounds, particularly if vegetation is abundant. Clearing shorelines reduces suitable breeding grounds and the mosquito population. There are thus steps that can be taken to attenuate the transmission of malaria after dams are built.

TABLE 1

SUMMARY STATISTICS

Notes:

Deaths/100,000 report the average number of malaria deaths per 100,000. Morbidity/10,000 reports the average number of reported malaria cases. Average monthly temperatures are reported in tenths of a degree Fahrenheit. Average monthly precipitation is reported in hundredths of an inch.

Sources:

See the Appendix.

The Tennessee Valley Authority

In May of 1933 President Franklin Delano Roosevelt created the Tennessee Valley Authority. It was chartered as a federal corporation that would assume control of the Tennessee River and its tributaries and promote development in the South. The charter announced that the TVA had a general mandate to improve the economy in the Tennessee Valley. The first director of the TVA, Arthur Morgan, quickly established plans for a series of dams and reservoirs to line the Tennessee River and its major tributaries. The plans expanded an existing U.S. Army Corps of Engineering (USACE) project, Wilson Dam in Muscle Shoals, Alabama and existing Army plans for dams at Cove Creek (Norris Dam) and Wheeler Reservoir. In its first six years, the TVA grew rapidly. By 1939 the TVA had received over $3.68 billion in federal appropriations. By 1951 cumulative appropriations exceeded $14.2 billion. Most of this money was spent on dams and related expenses including land purchases, transmission lines, and electric generation plants.

These projects also raised concerns because they impounded large bodies of water. It was estimated that dams' system of lakes and reservoirs would produce a shoreline 10,000 miles in length covering 600,000 acres of water. This vast shoreline created a large breeding ground for mosquitoes (Derryberry and Gartrell 1952). From the beginning, the TVA tried to mitigate malaria without heavy use of insecticides including introducing natural predators of mosquito larvae, destroying their habitat through drainage and periodic water fluctuations, brush clearing, larvacides, oiling, and eventually DDT (Gartrell and Ludvik 1954, TVA 1945).

Perhaps the TVA's largest effort came through the nearly costless plan to vary the water levels in the reservoirs. During the winter and early spring months, the TVA stored winter rain water at dams located at the top of the watershed. As the mosquito breeding season approached, the TVA released the stored water into downstream reservoirs, raising the water level a few inches. The surge flooded potential breeding grounds for mosquitoes. The following week, the water level in the dam was lowered exposing the mosquito larvae to natural predators. The water that was released was used to flood the shoreline of the next reservoir downstream, thus reducing its infestation. Water levels fluctuated throughout the entire system until the end of the breeding season.

The TVA's efforts to reduce malaria were the largest in the United States in terms of numbers of engineers and experts. As malaria rates began to decline, the TVA was praised for its strong efforts. The Centers for Disease Control (CDC) lists the TVA's accomplishments on its malaria history website.

It is clear that prevalence fell off, but that does imply that the TVA's actions were responsible because other agencies United States Public Health Service (USPHS), Works Progress Administration (WPA), and Malaria Control in War Areas (MCWA). Moreover, the net TVA impact is hard to predict because its action both worsened malaria incidence (dams increased sitting water) and reduced it (via vector control efforts). These factors help to motivate my effort to carefully estimate the net effects of a TVA dams on malaria rates.

Baseline Empirical Model of Malaria Rate

Let us start with a simple specification:

Where M_it is either the mortality rate per 100,000 people or the morbidity rate per 10,000 people in county i in year t. The vector C_i represents a set of county fixed effects to control for unobservable characteristics such as mountains, altitude, latitude, and longitude. Y_i is a vector of year fixed effects that seek to absorb things like region-wide epidemics and widespread shocks affecting the malaria rate.Footnote ³TVA_it is an indicator that equals 1 if county i in year t includes at least part of a completed TVA reservoir. Climate_it includes monthly average temperatures, monthly precipitation, and an interaction term of the two variables. X_it is a vector of time-varying county-level variables: the presence of a County Health Organization (CHO) to control for changes in access to health care, population density, percent black, and the urbanization rate.

This estimation strategy assumes that malaria was exogenous to and uncorrelated with dams locations—an assumption we will evaluate later. The model also assumes that the trend of disease rates did not differ between counties that received dams and those that did not. Figure 1 evaluates this assumption by showing the average mortality rate in Alabama from 1916–1933 by eventual TVA dam status.Footnote ⁴ Counties in black represent counties in Alabama that never received a TVA dam and counties in grey are those that eventually obtained a TVA dam. Between 1916 and 1933, prior to the creation of the TVA, there is no difference in the mortality rate between the two types of counties.

The impact of a dam can be seen in Figure 2, which presents a time series of the malaria morbidity rate for two adjacent counties, Limestone, located on Wheeler Reservoir and Madison, which is not on the reservoir. The first vertical line in Figure 2 denotes the beginning of construction, the second denotes its end. Prior to construction, the two counties had similar morbidity. However, the beginning of construction at Wheeler coincided with a major increase in the malaria morbidity rate in Limestone County. The increase in Limestone County persists through the completion of construction and remains higher than the initial rate for several years to follow. Madison County sees neither the rise nor the fall in malarial incidence. This comparison is just one example of the kind of differences between counties that help identify the effect of the TVA.Footnote ⁵ While there is a spike following initial construction, I cannot identify systematically the year in which the reservoir began to fill. Therefore the TVA variable is restricted to counties being located on completed reservoirs. Restricting the indicator in this way will bias the coefficient on the impact of the TVA (β₁) downward. Other effects will also present potential biases on β₁ Dams potentially lead to increases in wealth which in turn could lead to a decrease in the probability of contracting malaria. Dams also change the distribution of the population within the county by removing the population that was living close to the river. Both the income and population removal effects would lead to downward bias of β₁. One further complication in the empirical specification is that humans are mobile, possibly leading to spillovers in neighboring regions. This spatial aspect will be examined directly in the robustness checks with the use of spatially weighted regressions.Footnote ⁶

FIGURE 1

MALARIA MORTALITY IN ALABAMA, 1916–1935

Sources: Alabama State Board of Health (1916–1933).

FIGURE 2

MALARIA MORBIDITY NEAR WHEELER RESERVOIR

Sources: Alabama State Board of Health (1925–1951).

Is the TVA Exogenous to Malaria?

The underlying assumption of the empirical model outlined above is that the dams are randomly located with respect to malaria conditional on a variety of covariates. It is possible that dams are placed in areas that are prone to malaria—either because the TVA targeted these areas for their malarial levels or because the areas geologically most favored for dam building had high malaria incidence. However, the evidence suggests that malaria was not a concern when the TVA located its reservoirs. First, following the great flood of the Mississippi River in 1926/27, the USACE was commissioned to develop a strategy to prevent any recurrence of such disasters. The Corps of Engineers developed a plan to dam the Tennessee River so as to reduce the flood stage waters on the Mississippi River by one foot. This report detailed where dams needed to be built both for flood control and to provide a 9-foot navigation channel from Knoxville, TN to Paducah, KY where the Tennessee runs into the Ohio.Footnote ⁷ The 1930 USACE report included the engineering drawings for Norris Dam and Wheeler Dam, which TVA engineers used to construct those dams. Other TVA dams also relied on the locations outlined in that report. Although there is plenty of evidence that flood control and navigation played important roles in the placement of dams, there is no such evidence that engineers considered malaria.

TABLE 2

CHARACTERISTICS CORRELATED WITH DAM PLACEMENT AND MALARIA

* p < .10, ** p < .05, *** p < .01.

Notes:

River sizes are classified according to the definitions used in Fishback, Haines, and Kantor (2006). Elevation ranges are calculated by subtracting the maximum and minimum elevation in each county. The elevation categories are then classified as Low Elevation – below the 25th percentile of the sample, Medium Elevation – in the middle 50 percent of the sample, and Steep Elevation – in the top 25 percentile of the sample. Rainfall is the annual rainfall in the county.

Sources:

See the Appendix.

More importantly, the dams were randomly located with respect to pre-TVA malaria rates. Table 2, column 1 shows the results of a linear probability model that estimates the probability of obtaining a TVA reservoir by 1951 on, the presence of various sized rivers, elevation changes within the county, interactions between rivers and land gradients, and the malaria rate in 1930. These results show that dams were placed in areas with medium to large sized rivers and a steep land gradient. These results are consistent with the placement of dams designed to store water and produce electricity. They require steep changes in elevation to generate electricity, or a natural basin formed by the narrowing of the river through mountainous terrain. The results also show that the malaria rate in 1930 is not correlated with the placement of TVA dams.

While the placement of TVA dams is not correlated with malaria, it may be the case that features of the topography where dams are placed are conducive to more malaria. To address this concern, I regress the 1930 malaria rate on geographic characteristics, such as elevation changes, river flow, rainfall, and a full set of interactions between geographic characteristics. The results, presented in Table 2, column 2, show that malaria is correlated with areas that have a relatively flat land gradient, do not have moving water from medium to large sized rivers, and receive large amounts of rainfall. These results suggest that malaria is most prone in areas that are least suitable for large hydroelectric and storage reservoirs. The primary empirical model includes county-level fixed effects to take into account geographic features which do not change over time. However, if there is any correlation remaining, the results will downward bias the negative effect of the TVA on malaria's incidence.

One further concern with the placement of dams is that the USACE placed them in counties with desirable topographic characteristics and where it would be easy to assemble the land the reservoir would flood. If farmers abandoned their land immediately after it was purchased by the TVA, then there would be a large increase of idle agricultural land, which would likely raise malaria's prevalence (Humphreys 2001). Kitchens (2012) evaluates the TVA's use of eminent domain to assemble the properties for the reservoirs. He finds survey evidence that the Authority allowed farmers to stay on the land after it had been purchased. The TVA's typical policy allowed farmers to stay until their piece of property was going to be flooded, limiting the time that land was idle and, consequently, any increase in malaria.

Malaria Prevalence Data

Data has been collected from a variety of state and federal sources for every county in Alabama and Tennessee for the years 1926 through 1951 inclusive. Each year consists of 162 county observations over the 26 years which results in 4,212 county-year observations. Morbidity and mortality for Alabama and Tennessee come from state public health department reports and from the Vital Statistics of the United States.Footnote ⁸ These state-level reports provide the best insight into the prevalence of disease; however they also have many shortcomings. The earliest state health reports do not report malaria deaths or illnesses by race, and the only data pertaining to the age distribution of deaths is available at the state level in a limited number of years. Nevertheless, the county-level statistics are the best measures available for the time period and were used by policymakers to make decisions about funding and control efforts. For example, in 1945, when state health agencies first allocated DDT to counties, they did so based on average malaria mortality rates from 1938 to 1942 (Alabama State Board of Health 1946, p. 217). While there are some instances of more accurate blood smear surveys in elementary schools, these are few and far between, making county level statistics the most reliable information about malaria during the period. The malaria rate data allows malaria to be measured directly rather than indirectly as in past studies that construct an index of malaria prevalence from climate and geographic variables as in Sok Chul Hong (2007). It should also be noted that the reported rates may actually be a lower-bound estimate of the true prevalence in a county. Reporting of illness only occurred if a person was attended to by a physician. Contemporary malaria experts believed that the true prevalence was between 200 and 400 percent higher than reported rates.Footnote ⁹ Throughout our period, malaria was in steady decline. It was nearly eradicated by 1950. Figure 3 shows the spatial distribution of malaria mortality in Alabama and Tennessee in from 1930 to 1950. Deaths were heavily concentrated in the warm climates of South Alabama and in the wet regions of the Mississippi River in West Tennessee. The morbidity patterns are very similar.

I use monthly weather variables to control for random weather shocks that would affect malaria rates. To control for this, I have collected the Historic Climatology Networks Monthly Weather data from 1895–2009. This data contains the monthly average, maximum, and minimum temperatures as well as precipitation data collected at each weather station throughout the country. County-level observations were constructed through a triangular interpolation method.Footnote ¹⁰

FIGURE 3

SPATIAL PATTERN OF MALARIA MORTALITY

Note: Counties shaded in white represent zero deaths in the county. Counties in solid black represent over 12 deaths per 100,000.

TVA Annual Reports to Congress provided the locations of reservoirs. By the end of the sample period, 25/162 counties have a TVA reservoir located in their county. Figure 4 show the expansion of treated counties over time as new reservoirs were constructed by the TVA. The majority of counties are in eastern Tennessee and North Alabama, areas which traditionally had low malaria rates.

To proxy for the water level fluctuations from upstream storage reservoirs downstream through the system, I have created a variable that counts the number of reservoirs located upstream from each reservoir county. Upstream dams are necessary if engineers are to artificially alter the water levels so as to reduce the prevalence of malaria. Reservoirs further downstream are likely to gain the most due to increased control over the water level and flow rates through the reservoir.

FIGURE 4

EXPANSION OF TVA RESERVOIRS

Note: Counties shaded in black represent counties that lay on a TVA reservoir in the given year.

In both Alabama and Tennessee, the State Board of Health had offices in each county. Alabama had started to establish local offices in 1919, and by 1937 it a Board of Health office in every county. These offices aimed to fight a variety of diseases malaria, typhoid, syphilis, and other communicable diseases were at the forefront of importance for the local authorities. But they also were tasked with reducing infant mortality. The opening of a health agency is likely endogenous to malaria, wealth, and other measures of health; however it can be shown that the opening of a CHO and a TVA reservoir are uncorrelated, thus including the presence of a CHO will not bias the estimate for the TVA.

Other New Deal agencies also played a crucial role in the fight against malaria. Kitchens (2012b) shows that work by the WPA in Georgia led to large reductions in malaria during the 1930s. To account for ongoing work by other agencies, I include a set of three variables: New Feet, Old Feet, and Acres Excavated. New Feet refers to the number of linear feet of concrete lined drainage ditch constructed in the county-year. Old feet refers to the number of linear feet of drainage ditches which were cleaned, and Excavated refers number of acres of ponds or swamps filled with dirt or gravel. These measures were reported in Alabama's State Board of Health reported annual progress in counties where federal relief agency labor was used. Margaret Humphreys (2001) suggest that shoddy construction of drainage ditches and poor maintenance of ditches often led to increases in sitting water from sitting debris, whereas filling in swamp land with dirt or filling in ponds reduces the sitting water problem.

Population characteristics are compiled from the Census of Population and from the State Vital Statistics reports. Combining the population and area data provides a measure of the population density per square mile. In the early years of the malaria campaigns, state agencies focused on urban areas, recognizing that a large portion of the population could be affected by malaria fighting efforts. Because malaria is transferred from human to human through mosquito bites, residents of urban counties may be more susceptible to infection. Census data are also be used to determine the percentage of the population that was African American. There are two arguments assessing that malaria's impact might have been racially differentiated. The first is that sickle cell anemia, a genetic mutation concentrated among African Americans, provided resistance to infection and thus that group should have a lower infection rate (Allison 1954; Aidoo 2002). The second argument focuses on the poverty of southern blacks. Blacks may have to live on land closer to swamps and mosquito breeding grounds due to their low income status in the South, putting them at higher risk for contracting malaria (Humphreys 2001). It is unclear if sickle cell or poverty will play the dominating role.