Banking Development and Mortality Julio J. Elas and George T. McCandless Subgerencia General de Investigaciones Econmicas Banco Central de la Repœblica Argentina October 15, 2009 Abstract We consider both the theoretical and empirical issues of the e/ects of banks on health in an economy. We measure health by adult mortality rate. We present a model where people survive longer in an economy with money and banks than they do in an otherwise indentical economy with money but without bank. Our empirical results using cross country data show that the relationship between adult mortality rate and banking development, as measured by M2/GDP, is negative and signicant, even when other variables, such as income per capita, health expenditure per capita, education, are held constant. Este trabajo considera aspectos tericos y empricos de los efectos de los bancos sobre la salud en una economa. Utilizamos como indicador de la salud en la economa la tasa de mortalidad adulta. Presentamos un modelo en donde las personas viven mÆs en una economa con dinero y bancos en comparacin con una economa idØntica con dinero pero sin bancos. Nuestros resultados empricos utilizando datos de pases muestran que la relacin entre la tasa de mortalidad y el desarrollo del sistema bancario es negativa y signicativa, aœn controlando por ingreso y gasto en salud per capita, educacin y otras variables. JEL Codes: I1, G2 1 Introduction Banks provide a number of services to an economy. They provide households with savings and liquidity services. Bank deposits provide smaller savings with a safe place to save and the deposit contract normally allows depositors to withdraw their funds on demand. Some fraction of deposits are lent to rms, frequently for working capital. Banks provide the possibility of households to increase their savings so they can better meet the costs of emergencies (of interest here, medical emergencies) and these increased savings can reduce the cost of nancing working capital for rms and increase output. One of the most important types of emergencies that households face are health emergencies. In many countries, there is a correlation between the 1
Transcript
Banking Development and Mortality
Julio J. Elías and George T. McCandlessSubgerencia General de Investigaciones Económicas
Banco Central de la República Argentina
October 15, 2009
Abstract
We consider both the theoretical and empirical issues of the e¤ects ofbanks on health in an economy. We measure health by adult mortalityrate. We present a model where people survive longer in an economywith money and banks than they do in an otherwise indentical economywith money but without bank. Our empirical results using cross countrydata show that the relationship between adult mortality rate and bankingdevelopment, as measured by M2/GDP, is negative and signi�cant, evenwhen other variables, such as income per capita, health expenditure percapita, education, are held constant.
Este trabajo considera aspectos teóricos y empíricos de los efectos delos bancos sobre la salud en una economía. Utilizamos como indicadorde la salud en la economía la tasa de mortalidad adulta. Presentamosun modelo en donde las personas viven más en una economía con dineroy bancos en comparación con una economía idéntica con dinero pero sinbancos. Nuestros resultados empíricos utilizando datos de países muestranque la relación entre la tasa de mortalidad y el desarrollo del sistemabancario es negativa y signi�cativa, aún controlando por ingreso y gastoen salud per capita, educación y otras variables.
JEL Codes: I1, G2
1 Introduction
Banks provide a number of services to an economy. They provide householdswith savings and liquidity services. Bank deposits provide smaller savingswith a safe place to save and the deposit contract normally allows depositors towithdraw their funds on demand. Some fraction of deposits are lent to �rms,frequently for working capital. Banks provide the possibility of householdsto increase their savings so they can better meet the costs of emergencies (ofinterest here, medical emergencies) and these increased savings can reduce thecost of �nancing working capital for �rms and increase output.One of the most important types of emergencies that households face are
health emergencies. In many countries, there is a correlation between the
1
amount of health care services that a household can �nance and the care thata sick member of the household receives. This can be true even in economieswith extended public health care services. Expensive medical procedures ortests are rationed, one way or another, and private funds often permit quickeror easier access to such services. Quicker or easier access can result in bettermedical outcomes.If banks can improve the ability of households to access liquidity, they may
be able to improve the outcomes of health emergencies and therefore reducemortality. That is the issue we study in this paper. The question is whethereconomies with better developed banking systems have lower mortality rates,even after controlling for other variables. We provide a model where addingbanks reduce mortality (and increases life expectancy) and provide evidenceacross countries that supports the results of that model.Per capita income and mortality rates are highly correlated across countries.
So are the size of the �nancial system and per capita income. Yet, controllingfor per capita income and other variables, we �nd a strongly signi�cant negativemarginal correlation between adult mortality and M2/GDP, our measure of thesize of the �nancial system.We provide a model where adding a banking system can increase output
and reduce mortality. To make this result interesting, the baseline model hascash in advance money and households hold money for consumption purposesand for precautionary reasons. The baseline model is compared to one withDiamond-Dybvig style banks where deposits can be withdrawn in-period if thehousehold confronts a health emergency.In a Diamond-Dybvig bank, some fraction of the individuals are assumed
to confront a liquidity need. Here the liquidity need is made explicit andis endogenous. Some fraction of the households su¤er a health emergencyeach period. The health emergency is life threatening and spending additionalresources on confronting the emergency reduces the risk of death. The amountto be saved and to be spent on emergencies is a household decision that dependson the wealth of the economy and the institutions available. With this modelwe can say something about the welfare e¤ects of deposit taking banks thatput excess precautionary savings to productive use and about how part of thesewelfare e¤ects can be manifested in changes in output and in life expectancy.For the comparison to be interesting, the economy needs a number of fea-
tures. In the version without banks, both households and �rms face cash inadvance constraints. Households hold money for normal consumption purposesand to cover random large liquidity needs. The liquidity needs used here canbe thought of as a medical emergency where medical services need to be hiredand the more medical services purchased the higher the probability of survivingthe emergency. The advantage of using this kind of service is that the costs ofthe service automatically adjust to the cost of hiring labor. Firms need to hireand pay labor before they sell their goods. They need either saved or borrowedmoney to meet their wage bill.In the economy without banks, a substantial fraction of the money stock
does not participate in transactions in each period. Some funds are held for
2
precautionary purposes and are not all spent by the households that do notexperience an emergency. The amount of money held by each family is equalto the cost of consumption goods and emergency services for a family thatexperiences an emergency. Those who do not face an emergency have redundantcash. In the economy without banks, �rms hold cash between periods to covertheir wage bill. With banks, the excess liquidity of the households can be lentto the �rms for their working capital needs. One of the main activities ofcommercial banks is to use excess household liquidity to make short term loansto �rms. That is what banks do in this paper.The banks that are added to this economy are a very simple version of
a Diamond-Dybvig bank. They are one period banks, lending after both theemergency and technology shocks are realized, so they bear no risk. Householdshold money from the previous period. Some households use part of that moneyfor their emergency and the rest deposit the money in the bank. The bank thenlends to �rms for working capital. In a stationary state without money issue,the gross interest rate that banks can o¤er on deposits can�t go below one, sothere can be cases where not all of the money that is deposited in the bank getslent out to the �rms.An interesting set of results come from this model. Introducing banks into
the economy tends to (but does not always) increase output, increase consump-tion of goods by both those with the emergency and those without, increasethe hiring of labor to produce more emergency service, increase survival rates(and life expectancy), and generate a jump in the price level. In addition, realwages go up. Under the best conditions, introducing banks raises the returnthat households get on holding money for precautionary needs and providesfunds to the �rms at an interest rate lower than the implicit interest rate thatcomes from the �rms discount rate. The reduction in the interest rate paidby �rms means that they hire more labor and drives up the wage rate. Thismeans that the cost of medical services has gone up and whether more or lessemergency medical services are hired depends on whether the income e¤ect ofthe higher wages dominates the substitution e¤ect caused by labor becomingrelatively more expensive. Prices are higher with banks because more of themoney stock is involved in transactions in each period (some or all of the pre-cautionary savings that is not used for a medical emergency is now used by the�rms to pay wages).Mortality declines and output increases when banks are introduced in many
versions of the model. While there is a substantial literature on the importanceof precationary savings1 and the relationship between precationary savings andhealth, the literature has not highlighted the importance of the banking systemfor household health. With banks, the returns on precautionary savings arehigher and so the costs of health protection are lower. This has important policyimplications with respect to banking: making banking services (in our case,deposit services) available to households who do not now have it increases theirability to accumulate precautionary savings. These additional precautionary
1For example, Gourinchas and Parker [3] and Carrol and Samwick [1].
3
savings translate into lower mortality.
2 An economy without banks
We begin by constructing the economy without banks and then add banks tothat economy2 . Households and �rms face cash in advance constraints. Theconstraint holds for consumption and emergency medical purchases in the caseof households and for the wage bill in the case of �rms. Without banks, bothhouseholds and �rms need to carry money over from the previous period. Inthis economy, not all money will be used for payments in each period since thehouseholds who do not experience a medical emergency will have precautionarymoney holdings that they do not use.
2.1 Households
There are a unit mass of individuals in this economy. A fraction � (1� p (hxt )) ofthem die each period and are replace by and equal number of live but otherwiseidentical individuals who inherit their wealth of kt+1+
mt+1
Pt. The probability of
surviving the medical emergency, p (hxt ), is determined by the amount of medicalservices, hxt , that a household hires. The workers who provide medical servicesreceive the same wage as workers who produce goods.At the beginning of each period, a household discovers if it has a medical
emergency or not. With probability 1 � � a household does not require theemergency liquidity (nl) and faces the decision problem
Vnl (kt;mt) = max�u�cnlt ; h
nlt
�+ Et� ((1� �)Vnl (kt+1;mt+1) + �Vl (kt+1;mt+1))
�subject to
kt+1 +mt+1
Pt= wth
nt + rtkt +
nlt + �t + (1� �) kt +
mt
Pt� cnlt
and the cash in advance constraint
cnlt � mt
Pt
Here, kt is the capital carried over from the previous period, mt is the moneycarried over, cnlt is the goods consumption, hnlt is the labor supplied, �t arelump sum dividend payments from the pro�ts of the �rms, and nlt is a lumpsum tax or transfer that will make all surviving families have the same wealthat the end of each period. The depreciation rate is �, the wage rate is wt, therental rate on capital is rt and the price level is Pt.With probability � a household has to �nance a medical expenditure that
determines the probability that they will survive to the next period. Thedecision problem of those with liquidity needs (l) is
Vl (kt;mt) = max�u�clt; h
lt
�+ p (hxt )Et� ((1� �)Vnl (kt+1;mt+1) + �Vl (kt+1;mt+1))
�2The model here is based on McCandless [6].
4
subject to the budget constraint
kt+1 +mt+1
Pt= wth
lt + rtkt +
lt + �t + (1� �) kt +
mt
Pt� clt � wthxt
and the cash in advance constraint
clt + wthxt =
mt
Pt:
This cash in advance constraint says that the household will pay wthxt for med-ical services and will still consume clt. The amount of medical services they hireis monotonically related to the probability that they will survive into the nextperiod.To keep the model simple (and be able to aggregate the results), we add a
lump sum transfers program so that kt+1 +mt+1
Ptis the same whether one has
a liquidity demand (and lives) or not. Lump sum taxes for those who do nothave liquidity demands are nlt and the lump sum transfer to those who do itis lt. The transfer program has a balanced budget, so
0 = � lt + (1� �) nlt :
Since the probability of death in any period is � (1� p (hxt )), life expectancyof a person alive at the beginning of period t (before the liquidity need is re-vealed) is
1Xi=1
i�1� �
�1� p
�hxt+i�1
���i;
which, if one is in a stationary state, hxt+i�1 = hx, is
1Xi=1
i (1� � (1� p (hx)))i = 1� � (1� p (hx))[� (1� p (hx))]2
since � (1� p (hx)) is strictly between 0 and 1.
2.2 Production
There is a unit mass of identical, competitive �rms. The goods production sideof the economy can be expressed by the Cobb-Douglas production function
Yt = AtK�tH
1��t
where the equilibrium conditions for capital and labor are
Kt =
Z 1
0
ktdi;
and
Ht =
Z �
0
hltdi+
Z 1
�
hnlt di�Z �
0
hxt di;
5
and where At is the time t technology level.Firms have a cash in advance constraint in that they need to hold cash from
the previous period in order to cover their wage bill. De�ne mft as the money
that a �rm has carried over from period t� 1. LetR 10mft =Mf
t . The budgetconstraint of the �rms is
�t = Yt � wtHt � rtKt +Mft
Pt�Mft+1
Pt
subject to the cash in advance constraint
wtHt �Mft
Pt:
Firm managers maximize
Et
1Xi=0
�i�t+i;
and if the rate of gross in�ation is not less than �, the cash in advance constraintholds with equality so that
wtHt =Mft
Pt:
In a competitive economy, because of the e¤ects of having to hold moneyover from the previous period, pro�ts will be
�t = Yt � rtKt �Mft+1
Pt:
Using the �rst order conditions on rentals, we have
�t = Yt � �AtK�tH
1��t +
Mft+1
Pt:
The conditions for rentals is
1
�= Et
(1� �)At+1K�
t+1H��t+1
wt+1
!PtPt+1
:
2.3 Equilibrium conditions
All of the non-liquidity constrained households are alike as are the liquidityconstrained households. That means that
Cnlt = cnlt
andClt = clt:
6
The insurance plan means that
Kt+1 = kt+1
andMht+1 = mt+1;
since both the liquidity constrained and the non-liquidity constrained end upwith the same wealth and will allocate it in the same mannerMarket clearing conditions in each period for capital and labor are
Kt =
Z 1
0
ktdi;
and, de�ningHnlt = hnlt ;
H lt = hlt
andHxt = hxt ;
labor supplied to production is
Ht = (1� �)Hnlt + �H l
t � �Hxt :
De�ne the aggregate money held by the households into period t+ 1 as
Mht+1 =
Z 1
0
mt+1di:
The total money held by the �rms into period t+1 is Mft+1. A constant money
stock3 , M , is equal toM =Mh
t+1 +Mft+1:
As mentioned above, the zero pro�t condition for the insurance plan is
0 = � lt + (1� �) nlt :
2.4 Stationary states
It is possible to �nd the value of the value functions in a stationary state. Byimposing the stationary state conditions that kt;mt are constant through time,we know that
Vi = Vi (kt;mt) = Vi (kt+1;mt+1)
for both i = l and i = nl. In addition, because of the insurance program, theliquidity constrained that survive, the new households that replace the liquidityconstrained who die, the the non-liquidity constrained have the same stock of
3We are not considering the e¤ects of in�ation in this paper.
7
capital and the same money holdings. The discounted value of lifetime utilityin a stationary state can be written as
Vl =u�Cl;H l
�1� p(Hx)��
(1��(1��))
+p (Hx)� (1� �)u
�Cnl;Hnl
�h1� p(Hx)��
(1��(1��))
i(1� � (1� �))
(1)
for the liquidity constrained and
Vnl =u�Cnl;Hnl
�(1� � (1� �)) +
��
(1� � (1� �))Vl (2)
for those who do not face the constraint. The other �rst order conditions seethat the values of Cl;H l and Cnl;Hnl are those which meet the conditions fora maximum.For our example economy, the sub-utility function we use is
u�cit; h
it
�=
�cit�1�'
1� ' + b
�1� hit
�1�'1� ' ;
for i = l; nl, with 0 < ' < 1.The full set of 18 stationary state variables aren
The equations for the stationary state version of the economy without banksare given in equations 1 to ??.For an economy with � = :9; ' = :8; b = 1:2; a = 4; � = :1; � = :1; � =
:4;M = 1 and for a set of values for A = f1; 2; 3; 4g, the stationary state valuesare shown in Table 1. Real GDP is calculated by adding the real value ofgoods output to the real value of emergency services, GDP = Y + wHx.As might be expected, economies with higher levels of technology have higher
output, capital stocks, consumption in all states, and wages. In addition,economies with higher levels of technology work more, hire more labor for emer-gencies, have a higher probability of surviving the emergency (and therefore alonger expected lifetime), use relatively more money in the production processand use relatively less money by the households.
3 Adding banks
We add a simple, Diamond-Dybvig style bank to the previous model. In eachperiod, those who do not have a medical emergency deposit the money they arenot going to use for consumption into a bank. The bank lends these funds tothe �rms to help cover the wage bill. The banks are mutual so that all interestpaid by the �rms is passed along to the households.
Households maximize the same discounted utility function as in the previoussection. However, the budget and cash in advance constraints are di¤erent.The households without emergencies maximize subject to the budget constraint
cnlt + kt+1 +mt+1
Pt= wth
nlt + rtkt +
nlt + �t + (1� �) kt +
mt
Pt+�rht � 1
� ntPt
and the cash in advance constraint
cnlt +ntPt=mt
Pt:
Instead of holding excess money, they deposit all the money they are not usingfor consumption into the �nancial system and receive the gross return rht onthose deposits.Households with the emergency expenditures maximize subject to the budget
constraint
clt + wthxt + kt+1 +
mt+1
Pt= wth
lt + rtkt +
lt + �t + (1� �) kt +
mt
Pt
and the cash in advance constraint
clt + wthxt =
mt
Pt:
9
All of their cash holding is used to �nance consumption and the services thatthey pay for given the emergency. Since, as we will see, capital pays a higherreturn than bank deposits, households will hold only the amount of money theyneed to cover their desired expenditures during an emergency. In an importantand probably realistic way, households are credit constrained. In our economy,those who face a medical emergency cannot borrow from the banks to cover theexpenses of that emergency.There are some corner conditions here. If the expected interest rate paid
by the banks becomes less than rht = 1, all households will hold only money,since, with a constant money supply, the expected rate of return on money willnot fall below that amount
3.2 Banks
Banks take in the deposits of those who do not have emergencies and lend thesefunds to the �rms to cover all or part of their wage bill. Banks make no pro�ts(are mutuals) and lend at the same rate that they borrow from the depositors.Banks do not make loans to individuals who have emergencies. They only makeriskless in-period loans to �rms. Since only those without emergencies depositin the banks, total deposits available to the �rms are
Nt = (1� �)nt:
Banks lend these are the rate rft = rht . Banks lend all the deposits they receiveto the �rms.
3.3 Firms
If the interest rate on borrowing from the banks is less than 1=�, their costof holding money, the �rms will borrow as much from the banks as they can.This is Nt = (1� �)nt. The �rms will save from the previous period Mf
t tocover the expected di¤erence between their borrowings and their desired nominalexpenditure on labor. The aggregate cash in advance constraint for the �rmsis
Nt +Mft = PtwtHt:
The �rms are maximizing the value of the �rm
Et
1Xi=0
�i�t+i;
subject to the budget constraint
�t = Yt � wtHt � rtKt ��rft � 1
� NtPt+Mft
Pt�Mft+1
Pt
and the cash in advance constraint. The production function is as before,
Yt = AtK�tH
1��t :
10
First order conditions for the �rms are
(1� �)AtK�tH
��t = rft wt
�AtK��1t H1��
t = rt1
�� Etr
ft+1
PtPt+1
:
The last conditions is with inequality if Mft = 0, if the �rm can borrow from
the banks all of the funds it needs to �nance the wage bill. If it cannot borrowenough, then Mf
t > 0 and the condition is an equality (which implies that in astationary state without in�ation, rf = 1=�).
3.4 Equilibrium conditions
Most of the equilibrium conditions are the same as those in the economy withoutbanks. The major di¤erences are in the conditions for the banks, which weassume are competitive and therefore lend all the deposits they receive if theinterest rate that the �rms pay is greater than rft > 1.
3.5 Stationary state
We use the same sub-utility function and probability function as in the no bankeconomy.The full set of 20 stationary state variables aren
The two additional variables are the interest rate, rf , and bank deposits, N .Results for simulated economies with the same parameter values as for the no-bank economy and with A = 1; 2; 3, and 4 are given in Table 2.Compare the results here to those of Table 1. For the example economy
with banks, output is higher, expenditures on medical services are higher andtherefore the probability of surviving to the next period is higher. While notshown here, it can be the case that with banks, the additional savings pullsso much extra labor into the medical services industry that output of goodsdeclines.
4 Empirical Evidence on the Relationship be-tween Banking Development and Mortality
In most empirical studies on the determinants of health, education and incomeappear as the most important correlate of good health, regardless of whether
11
Table 2: Values for simulation in bank economyvariables A = 1 A = 2 A = 3 A = 4GDP 0:5460 1:9861 4:2191 7:1891Y 0:4889 1:7926 3:8244 6:5357Cnl 0:4226 1:5361 3:2612 5:5546Cl 0:1755 0:7582 1:7567 3:1645Hnl 0:3889 0:4441 0:4789 0:5044H l 0:3629 0:4223 0:4593 0:4863Hx 0:6331 0:6924 0:7261 0:7493H 0:3230 0:3727 0:4043 0:4277K 0:9107 3:3430 7:1365 12:2006w 0:9008 2:7946 5:4346 8:7210r 0:2148 0:2145 0:2144 0:2143P 1:3407 0:3713 0:1754 0:1031� 0 0 0 0
health is measured by mortality rates, morbidity rates, or self evaluation ofhealth status, and regardless of whether the units of observation are individualsor groups (see Grossman [4]). Some recent work has concentrated on the e¤ectsof health risk on saving behavior. Using data for Italy Japelli, Pista¤erri andWeber [8] found a positive relationship between health dispersion and precau-tionary savings. On the other hand, Starr-MacCluer [7] found mixed evidenceon the e¤ects of health risk on precautionary savings.The e¤ect of the banking sector, as a self protection mechanism, on mortality
rates has not been explored in previous studies. The goal of this section is toinvestigate empirically this relationship using cross country data on mortalityfrom the World Health Organization statistics. According to our results in thissection the association between banking system development, as measured byM2/GDP, and adult mortality is quite important.Before entering into the econometric analysis, it is worth noticing that sav-
ings for precautionary purposes at the household level, as modeled in previoussections, are empirically important. For instance, according to the 2007 Fed-eral Reserve Board�s Survey of Consumer Finances, the second most frequentlyreported motive for savings is liquidity related (with 32 percent of the house-holds), the �rst being retirement motives with 33.9 percent of the households.Liquidity-related reasons include �emergencies,� the possibilities of unemploy-ment and illness, and the need for ready money.The same survey also shows that families estimate at around 9.2 per cent
of their income the fraction of savings they need for emergencies and otherunexpected contingencies. For low income families this fraction is even larger,the desired precautionary savings represents 14 per cent of their income.The model developed in previous sections predicts a negative relationship
among mortality rate and the relative size of the banking system across economicunits, after controlling for income. In the model, households save some resourcesfor health precautionary motives, aside from the standard reasons for saving.The banking system allows households to save at lower costs, what encouragesavings (and banks) and, as a consequence, reduces household�s mortality.There are various ways to measure banking sector size and development.
The traditional indicator utilized for assessing the size and development of acountry�s banking sector is the ratio of M2 to GDP. Other measure used inthe literature includes the ratio of private credit to GDP, number of accounts,bank branches and ATMs per population. In order to make our sample as largeas possible, we use the ratio of M2 to GDP for the purposes of evaluating themodel. In Table 3 we provide the details about the data used in the analysis.
Figure 1 shows the correlation between adult mortality rate (probability ofdying between 15 to 60 years per 1000 population) and the size of the bankingsector in logs, as measured by M2 over GDP (without any additional controls).The �gure shows a strong negative association between both variables. Thecorrelation coe¢ cient between the two variables is -0.49. As the size of thebanking sector increases, mortality rate declines. However, taking into account
There are 150 observations. Numbers with * are signi�cant at 1% level.
Table 5: Baseline Relationship between Adult Mortality and M2/GDP acrosscountries - Average Mortality Rate by M2/GDP and GDP per Capita
M2/GDP Difference
GDP per Capita (PPP) US$Less than 36%
(Median) More than 36%All
Countries(1) (2) (2) – (1)
Less than 6757 (Median) 347.3 203.5 307.7 144.2*More than 6757 (Median) 204.6 122.5 147.8 82.1*All Countries 302.5 168.1 225.1 134.4*
There are 150 observations.
the well known relationships between income and both mortality rate (see Figure2) and �nancial development (see Table 4), the relationship plotted in Figure 1may not seems surprising.In Table 5, we take a crude look at the relationship between banking de-
velopment and mortality conditioning on income per capita by breaking downmortality rates according to countries income per capita and banking develop-ment (M2/GDP). The table provides an overview of the data and shows thatthe negative di¤erences in mortality rates according to the level of developmentof the banking sector are signi�cant among both high and low income countries.The di¤erence in mortality rates by banking development is larger for less devel-oped countries, but they are similar for both group, around 40%, in percentageterms.In what follows, using regression analysis we show that, consistent with the
model, the negative correlation between adult mortality rates and M2/GDPsurvives even after controlling for income and other variables that may be cor-related with both mortality and �nancial development. We also analyze thesensitivity of our results to di¤erent speci�cations, the choice of the dependentvariable and, to a lesser extent, the measure of banking development.Table 6 presents our baseline results. All regressions are run by OLS, with
the dependent variable the log of adult mortality rate. There are 7 columns,
16
Table 6: Cross Country OLS Regressions of ln of Adult mortality rates onM2/GDP and other variables
Robust standard errors are reported in parentheses. Numbers with ** signi�cant at 5%.Numbers with *** signi�cant at 10%.
corresponding to di¤erent speci�cations. Our interest is in the coe¢ cient ofM2/GDP, our measure of banking development. We control for the usual cor-relates included in health regressions including education, income and healthexpenditures, the main explanatory variables in health regressions. In somespeci�cations we also include a dummy variable for African countries. Thesample consists of 150 countries.The results show that the relationship between adult mortality rate and
M2/GDP is negative and signi�cant, even when other variables, such as incomeper capita, health expenditure per capita, education, are held constant. Incolumns 5, 6 and 7, we report OLS results of similar speci�cations but weightingby population. The magnitude of the e¤ect of M2/GDP on mortality increaseswhen we weight by population size.As a robustness check, in Table 7 we report regressions results using as de-
pendent variable mortality rates by gender and life expectancy instead of mor-tality rates for the population. As the table shows, the e¤ect of our measureof banking development, M2/GDP, is signi�cant at 1% level in most speci�ca-tions. One curious result of this analysis is that the e¤ect banking developmenton mortality rates is much stronger for males than for females. When we uselife expectancy as dependent variable, the coe¢ cient is not signi�cant withoutweighting for population.In the last column of Table 7, instead of using M2/GDP as explanatory
variable we use an index of banking development from the Financial Develop-ment Report of the World Economic Forum. Even tough the sample is reducedbecause of data availability, we still �nd a signi�cant negative e¤ect of bankdevelopment on mortality rates.
17
Table 7: Cross Country OLS Regressions of ln of Adult mortality rates by genderand ln of Life Expectancy on Banking Development and other variables
Robust standard errors are reported in parentheses. Numbers with ** signi�cant at 5%.Numbers with *** signi�cant at 10%.
Overall, the results presented in this section suggest that the relationshipbetween banking development and mortality is empirically important. Theseestimates suggest a statistically signi�cant negative e¤ect of banking develop-ment on mortality rate.
5 Conclusions
This paper considers both the theoretical and empirical issues of the e¤ectsof banks on health in an economy. We present a model and evidence thatsuggests that a more developed banking sector positively a¤ects the health ofthe population.The model predicts a negative relationship among mortality rate and the
relative size of the banking system across economic units, after controlling forincome. In the model, households save some resources for health precautionarymotives, aside from the standard reasons for saving. The banking system allowshouseholds to save at lower costs, what encourage savings (and banks) and, asa consequence, reduces household�s mortality.Using cross country data on countries mortality rates from WHO, we esti-
mate the relationship between M2/GDP and mortality rates. The results showthat the relationship between adult mortality rate and M2/GDP is negative andsigni�cant, even when other variables, such as income per capita, health expen-diture per capita, education, are held constant. Overall, the results presentedin this paper suggest that the relationship between banking development and
18
mortality is empirically important.According to our analysis, the existence and size banks are important for
health in a society including when output and other relevant variables are takeninto account. Our model suggests that regions with less access to bankingservices are likely to be less healthy. A direct implication of the analysis is thatbanking regulations that make it more costly to use banking services probablyare a¤ecting people�s health.
References
[1] Carroll, C. & Samwick, A. (1997), "The nature of precautionary wealth,"Journal of Monetary Economics, 40(1), p. 41-71.
[2] Diamond, D., and P. Dybvig (1983), "Bank runs, deposit insurance, andliquidity," Journal of Political Economy, 91(3), p.401-19.
[3] Gourinchas, P.-O., and J. Parker (2001), "Precautionary Savings: The Em-pirical Importance of Precautionary Savings," AEA Papers and Proceedings,May, 91(2), p.406-412.
[4] Grossman, Michael (2003), �Education and Nonmarket Outcomes.�Hand-book of the Economics of Education, edited by Erik Hanushek and FinisWelch. Amsterdam: North-Holland, Elsevier Science.
[5] Kiotaki, N., and J. Moore (1997), "Credit cycles," Journal of Political Econ-omy, 105(2), p. 211-248.
[6] McCandless, G.T. (2009), "On the social value of banks," Working paper,Banco Central de la República Argentina.
[7] Starr-McCluer, M. (1996), "Health insurance and precautionary saving,"American Economic Review, 86, pp. 285�295.
[8] Jappelli, T., L. Pistaferri & G. Weber (2006), "Health Care Quality, Eco-nomic Inequality, and Precautionary Saving," "Marco Fanno" Working Pa-pers 0020, Dipartimento di Scienze Economiche "Marco Fanno".
