Sobering Up: The Impact of the 1985-1988 Russian Anti-Alcohol Campaign on Child Health

Andreea Balan Cohen Tufts University

First Draft: June 2006 This Draft: July 18th, 2007

Abstract. This paper estimates the impact of parental alcohol consumption on child health by taking advantage of a unique shock to alcohol supply: the 1985 to 1988 alcohol prohibition campaign in Russia. This campaign was temporally short lived, and resulted in large amounts of exogenous geographic variation in its intensity and effectiveness. I construct a new data set that combines the Russian Longitudinal Monitoring Survey with regional data on alcohol consumption. I find significant improvements in child height, immunization rates, and chronic conditions among children born during prohibition who also lived in regions with effective antialcohol campaigns. I find no effect on children born either before or after prohibition. This confirms the effect of investments during a child’s first three years of life on long-term health measures, and demonstrates a potential positive effect of suppressing parental access to alcohol. Furthermore, evidence from vaccination rates suggests that the positive effect of prohibition on child health occurred through improvements in parental time, rather than income resources.

Keywords: child health; alcohol prohibition; fetal origin; Russia JEL classification : O12; I12; I38; J13; P36

Introduction Parental alcohol abuse affects millions of children worldwide. One out of ten American children and one out of eight European children live in a household with at least one alcohol dependent or alcohol-abusing parent (Huang et al.1996, Eurocare1998). The economics literature has mostly focused on the effects of parental alcohol consumption on child abuse and mental health (Jones et al.1999; Markowitz 2000; Grossman and Markowitz 1998, 2000; Chatterji and Markowitz 2001), but parental drinking can have a significant impact on other aspects of child health as well. In recent work, for instance, Bonu et al. (2004) document that children from Indian households that used tobacco or alcohol were more likely to have acute respiratory tract infection, more likely to be malnourished, and more likely to die before their first birthday.1 Despite the existence of a positive correlation between substance abuse by parents and adverse physical and mental health outcomes in children, establishing a causal relationship has proven difficult. The observed relationship may be causal if alcohol consumption has a direct impact on parenting ability or the amount of resources (both income and time) that parents invest in children. On the other hand, the relationship may be the result of unobserved factors that are correlated with both parental alcohol consumption and child outcomes, such as parental psychiatric disorder, stressful home environment, or living in a dangerous neighborhood. To control for these confounding factors, researchers have used (1) child and family specific fixed effects models, which control for unobserved heterogeneity at the level of the child, and parents’ family of birth, respectively; (2) instrumental variables methods, which use state alcohol prices and policies to identify parents’ alcohol consumption (Markowitz 2000, Grossman and Markowitz 1998, 2000, Chatterji and Markowitz 2000).

1

See Gmel and Rehm (2003) for an extensive review of the possible effects of alcohol consumption on child and relatives’ lives.

Even with these more careful approaches, however, some issues still remain. This paper extends this area of research by examining the impact of the 1985-1988 anti-alcohol campaign in Russia, which generated large alcohol consumption decreases that varied across regions. The primary data source is the Russian Longitudinal Monitoring Survey, a rich longitudinal data set on child and parental health, height and other individual and family characteristics, which is combined with regional alcohol consumption data. The contributions of this paper are several. First, I focus on physical measures (height, chronic health conditions and immunizations) rather than mental measures of child health. This diminishes the problem of certain confounding factors—such as genes and personality—being correlated with parental alcohol consumption, since these factors are much more likely to influence child mental outcomes rather than physical health. Second, by focusing on national rather than state (local) alcohol policy and on a time period when internal migration in Russia was restricted, the endogeneity of families’ location in response to changes in alcohol prices and programs is not an issue in the estimation. Third, I show not only that restrictive alcohol policies can have a large positive effect on child physical health, but also that this effect occurs even in heavy drinking environments. Finally, I present some evidence on the channels through which parental alcohol consumption affects child health. In particular, I show that, in Russia between 1985-1988, parental time inputs might have been more important contributors to child health than parental monetary investments. This paper also contributes to the literature on the effects of the 1985 to 1988 anti-alcohol campaign in Russia, and on the medium-term effect of prohibitions more generally. The effect of the Russian prohibition on (adult) health has been hotly debated. Some authors have argued that the prohibition was associated with dramatic decreases in adult mortality, as well as with reduced crime incidence (Nemtsov and Shkolnikov 1997). Other authors, however, have argued that the

beneficial health and social effects of the anti-alcohol campaign have been significantly overstated due to problems with both the official alcohol data and the mortality calculations (Treml 1997). This paper contributes to this literature by focusing on a health outcome other than adult mortality, and by combining unofficial alcohol data at a regional level with height and health data at the individual level to show that the prohibition did have positive (medium to longrun) effects on child health. The paper proceeds as follows. Section 1 provides a background on the alcohol campaign, and sections 2 and 3 describe the analytical framework and the data, respectively. In section 4, I discuss the empirical strategy, and in section 5 I present the results. Section 6 concludes.

1. Alcohol Consumption and the 1985 to 1988 Anti-Alcohol Campaign In the Soviet centrally planned economy, the state had a complete monopoly on the legal production, pricing, foreign trade, and distribution of alcohol. Since excise taxes and state profits from alcohol sales represented a large fraction of Soviet government revenues, it is perhaps not surprising that between 1960 and 1984, the sale and production of alcohol in the Soviet Union more than doubled, from 4.6 to 10.5 liters of pure alcohol per capita (see figure 1).2 In the 1980s, recorded alcohol consumption per capita in Russia was higher than alcohol consumption in most OECD countries.3 Alcohol consumption was rapidly becoming a serious societal problem: the age at which people started drinking was falling rapidly, an increasing number of women and children were becoming serious drinkers, and, in some cities, average consumption among working age adults was a bottle of vodka per day (White 1996).

2

10.5 liters of pure alcohol per capita is roughly the equivalent of 22 liters of 100-proof vodka per person per year. The country with the highest alcohol consumption per capita in 1990 was France (12.7 liters), but most other OECD countries had per capita alcohol consumption in the 5-9 (liters) range.

3

Prior to 1985, there had been some half-hearted attempts on the part of the Soviet government to address the issue of alcohol abuse. Two anti-alcohol resolutions of the Central Committee of the Communist Party had been released in 1958 and in 1972 under Brezhnev, and after 1982 some action was initiated by Andropov and Chernenko under the general heading of anti-social behavior. None of these measures had met with much success, however (Nemtsov and Shkolnikov 1997, McKee 1999, Richardson 1999).4 The anti-alcohol campaigns prior to 1985 had attempted to address the alcohol issue through public health education approaches, encouraging moderate drinking, the substitution of wine or beer for vodka, and increasing intolerance towards drunk-driving and drunkenness in the workplace. When Gorbachev succeeded Chernenko in 1985, however, these measures were rejected as too half-hearted, and replaced instead by an all-out war against alcohol.5 The anti-alcohol campaign was announced in April 1985, began being implemented in earnest in June 1985, and was ended in 1988. It included a wide array of measures: alcohol was banned at all official functions and in public places, particularly on trains, near industrial enterprises, and workplaces; party officials and managers who drank heavily were dismissed; alcohol prices were increased twice (in August 1985 and again in August 1986); the penalties for the production and sale of home-made samogon (moonshine) were raised and more strictly enforced (Ivanets and Lukomskaya 1990, McKee 1999, Tarschys 1993). Finally, and most importantly, state production and sale of alcohol was massively reduced; by 1987, the number of stores selling wine and vodka in Russia was 5 times lower than in 1984, and the agricultural acreage for wine grapes was 30 percent lower (Nemtsov and Shkolnikov 1997).

4

In 1984, following a call by Chernenko for stricter enforcement of the existing alcohol legislation, alcohol consumption finally started falling, though not by much (McKee 1999). 5 Within a few months of his designation as new Secretary General, Gorbachev became known as ‘Mineral Water Secretary’ due to his radical stance on drinking (Tarschys 1993).

These measures had a very strong and immediate impact. The queues at official alcohol outlets became as long as 3000 people each day, road traffic accidents and absenteeism from work due to alcohol-related causes decreased, and state receipts from alcohol sales plummeted.6 As can be seen from Figure 1, between 1985 and 1987, recorded alcohol consumption dropped 56 percent, from 8.8 to 3.9 liters of pure alcohol per capita. Furthermore, as Figure 2 shows, the consumption of all state-produced alcoholic beverages fell during this time period: the consumption of vodka and wine by 55 percent, and that of beer by 26 percent, respectively. Official alcohol consumption data figures do not include the consumption of homemade alcoholic beverages (samogon). Even before the start of the anti-alcohol campaign, samogon consumption was as high as 30 percent of the official alcohol consumption (Nemtsov and Shkolnikov 1997). The anti-alcohol campaign initially included severe penalties for the production and sale of samogon, but as a result of the loosening of political restrictions (‘glasnost’) in the late 1980s, the prosecution of minor law-breaking offenses, including alcoholrelated ones, declined over time. As a result, the anti-alcohol campaign was associated with an increase in samogon consumption, especially after June 1987 when first time convictions for home brewing of alcohol became a non-criminal offense (McKee 1999). Since the production and purchase of samogon in Soviet Russia was illegal, the quantification of samogon consumption is difficult. Between 1971 and 1989, the Soviet statistical agency, Goskomstat SSSR did produce estimates of samogon production, but these estimates suffer from some problems, especially after 1987.7 Nemtsov (1992) and Nemtsov and Nechaev (1991) also estimated actual alcohol consumption in Russia on the basis of the 6

State revenues from alcohol fell by 5 billion rubles between 1984 and 1985, and by 15.8 and 16.3 billion rubles, respectively, in the following years (McKee 1999). 7 The Goskomstat estimates of samogon are restricted to sugar-based samogon, and exclude samogon produced from other inputs, such as potatoes, grain, and fruits, as well as home-made wines and beers (see Treml 1997). The Goskomstat method of estimation broke down beginning in the 1988 due to an acute sugar shortage, and was consequently abandoned after 1989 (Nemtsov 1992, McKee 1999).

proportion of violent deaths involving the presence of alcohol in the blood. As can be seen from Figure 3, for the period 1983-1988 Nemtsov’s estimates are very similar to the sum of the Goskomstat estimates of samogon consumption and the official values of state alcohol sales, and they indicate a smaller—but nevertheless very substantial (26.5 percent)—drop in total alcohol consumption during the anti-alcohol campaign.8 By the early 1988, however, the campaign was weakening. The consumption of samogon as well as industrial alcohol was rising, Russian government finances were increasingly strained due to the absence of alcohol profits, and the anti-alcohol campaign was becoming increasingly unpopular. In the winter of 1988, Moscow authorities responded to the numerous complaints about the unavailability of vodka by increasing the number of outlets and trading hours, and in October 1988 the production of alcohol across the Soviet Union was increased so as to eliminate queues, effectively ending the anti-alcohol campaign (Tarschys 1993). Between 1988 and 1991, the backsliding in alcohol consumption was rather slow. By 1993, however, following the hyperinflation of the early 1990s—during which the price of alcohol decreased sharply relative to personal salaries and the general price index—alcohol consumption rose back to its pre-1985 level (see Figure 3).9

2. Analytical Framework Child health is a function of genetic endowments, in utero health, as well as nutrition and other forms of investment during childhood. Parental alcohol consumption can negatively affect physical child health in two primary ways: the first is through alcohol consumption by pregnant mothers, and the second is by diminishing the financial inputs and time available for childcare 8

The difference between Nemtsov’s and Goskomstat’s estimates becomes much larger during the 1988-1989 time period, when the deficit of sugar in the state trade biased the Goskomstat estimates of samogon production. 9 The general price index increased 1229 times between December 1992 and June 1994, whereas alcohol prices increased only 421 times during this time period. As a result, real alcohol consumption during the early 1990s increased sharply (Shkolnikov and Nemtsov 1997).

(Bonu et al.2004). In Russia before the prohibition, both of these latter effects were large. In 1984, for instance, alcohol consumption accounted for between 10 and 15 percent of Russian household budgets (Tarschys 1993). In addition, drinking was associated with diminished wages due to work absenteeism and unemployment, as well as with income losses due to the drinking parents’ increased morbidity and mortality (White 1996).10 Quantifying the magnitude of the impact of drinking on parental time resources is harder, but indirect evidence suggests that it was large as well. For the average alcoholic worker, for instance, the number of work days lost due to alcoholism was about 93 days a year (White 1996).11 Since weekends and holidays were also associated with alcohol excesses, the loss of parental time available to spend with children was probably even larger. Therefore, the decline in alcohol consumption during the prohibition had a large potential for improving child health outcomes. Its impact, however, was slightly mitigated by three factors: increases in alcohol prices, increases in the time required to purchase alcohol, and the substitution of more dangerous substances for alcohol. To begin with, during the prohibition alcohol prices were raised twice, by 25 percent in August 1985, and then again in August 1986 by a further 20 percent. The effect of the first price increase on household budgets was mitigated by decreases in the price of most foodstuffs and household items, but that of the second price hike was not (White 1996). Second, alcohol shortages led to long queues at the state supply stores, with people waiting as long as 10 hours in line to purchase alcohol.12 Finally, the restrictions on alcohol led to an increased consumption of moonshine, and also of (sometimes toxic) alcohol-containing substances such as cologne, glass cleaners, and glue, with potentially 10

Segal (1990) estimates that 30 percent of the Soviet labor force suffered from alcoholism. Similarly, a survey from a chemical plant from the early 1980s revealed that almost 25 percent of the workforce consisted of alcohol abusers (White 1996). 11 In some regions in the countryside, collective farm chairmen were able to find sober workers only during the first half of the day (White 1996). 12 A joke from the time period, for instance, has the bus driver announcing the liquor stop, and, three bus stops later, the end of the queue to the liquor store.

negative consequences for drinkers’ health (Treml 1997, White 1996).13 Since the impact of these three factors was largest in the last year of the prohibition, 1988, I will examine this time period separately to check for their relative importance.

3. Data Description As I discuss in greater detail in section 4, I estimate the long-run impact of prohibition on child health by taking advantage of the variation of prohibition across birth cohorts and regions. For this purpose, I combine child and parent data from round 5 of the Russian Longitudinal Monitoring Survey (RLMS) with a dataset on regional alcohol consumption. Direct information on child outcomes between 1985-1988 is not available, but since health during early childhood has significant effects on later life health (even as late as adulthood), I am able to measure the effect of the anti-alcohol campaign by focusing on child outcomes seven years after the end of prohibition.14 For these purposes, I use the RLMS survey, which contains very detailed income, family background, and health and anthropometric information for a nationally representative sample of 3,975 households (2397 children) interviewed in December 1994.15 Table 1 and 2 present summary statistics on the household controls and health outcomes used in this paper. Since chronic conditions later in life are associated with poor childhood nutrition and care (Barker 1995, Fogel and Costa 1997, Manton et al. 1997, Ravelli et al. 1998), I examine the impact of prohibition on the probability of the child reporting to be in good health, having had any chronic health problems, having been hospitalized, or having had chronic respiratory congestion. 13

According to a report to the Central Committee of the Communist Party (cited in Treml 1997), sales of certain types of alcoholbased glue increased from 760 tons in 1985 to 1,000 tons in 1987, and sales of glass cleaners increased from 6,500 to 7,400 tons in the same period. 14 See Haas 2007, and Kuh and Wadsworth 1993 for reviews of this literature of the “long arm of childhood”. 15 Although the RLMS survey has been conducted since 1992, data from rounds 1-4 (1992-1994) is not representative at the national level (see Zahoori et al. 1999 for details).

To control for the possibility that these health outcomes were determined by more recent income and nutrition shocks during the children’ lives (rather than prohibition), I perform placebo tests using as outcomes indicators for acute (rather than chronic) health conditions— coughing, sore throat, and diarrhea. In addition, I also use long run measures of health that are determined at specific ages during childhood, namely height and immunizations. Height has been shown in numerous studies to be a good proxy for early life health and development (Falkner and Tanner 1986, Floud et al. 1990), and the period in a child’s life between ages zero and three is considered critical to determining later life height, especially prior to adolescence (Beard and Blaser 2002).16 The reason for this is that the speed of growth and nutritional needs are greatest during this period, and so is the risk for poor parental care-giving and (growth-retarding) respiratory and gastrointestinal infections (Martorell et al. 1994).17 In order to compare height consistently across birth cohorts, I control for age using flexible (both parametric and nonparametric) functional forms, and I also construct standardized height for age (HFA) z-scores.18 Similarly, immunizations have specific age schedules for being administered. While in developed countries “catch-up” immunizations can be administered at older ages for children or adults who were not vaccinated at the recommended times, this was not the case in Russia prior to 1989, due to administrative and bureaucratic constraints. I match the data on child outcomes and parental and household characteristics with alcohol consumption data in the child’s region of birth for the time period 1981 to 1992. The alcohol consumption data series is described in Nemtsov (1992, 1997) and Nemtsov and

16

Height deficits accumulated during childhood cannot be erased by the growth spurt during adolescence, but they can be lessened somewhat (Martorell et al. 1994, see also the review in Case and Paxson 2006) 17 Older, more autonomous children seem to be better equipped to protect themselves against the effects of poor parenting (Martorell et al. 1994). 18

Height for age z-scores are standard deviations from the NCHS reference median, as suggested by the World Health Organization. A height-for-age z-score of -2.0, for instance, implies that the child is two standard

deviations below the median of the reference population.

Nechaev (1991). As discussed in section 1, this data refers to total alcohol consumption (samogon included), and is consistent with official alcohol sales and samogon estimates.19 Summary statistics for the alcohol data are provided in Table 1. To diminish the potential measurement error in this data (due to problems in measuring samogon consumption), I construct an indicator variable for the intensity of prohibition, which takes the value 1 if the change in alcohol consumption during prohibition compared to 1984 was above the median regional change. In section 5, I check the sensitivity of the results to alternate classifications of prohibition intensity across regions.

4. Empirical Strategy To estimate the impact of anti-alcohol campaign on long run child health, I take advantage of the exogenous variation of the effects of prohibition across regions and birth cohorts. 4.1. Identification (I): Cohort Variation in Exposure to Prohibition

As discussed in section 1, the alcohol consumption during the prohibition period was much lower compared to that in the early 1980s, and also compared to the post-prohibition years. Children born during the prohibition years, therefore, had very different levels of exposure to parental alcohol consumption compared to children born after 1988. Children who spent a larger fraction of their first three years of life under prohibition also had the potential to experience larger changes in outcomes compared to other cohorts, due to the crucial impact of this postnatal time period on children’s long run health. Finally, the effect of prohibition on the cohorts born prior to 1985 is unclear, because they had spent the later part of their childhood under prohibition, and the early part of it under a “heavy” alcohol consumption regimen.20 It is important to note here that since the prohibition did not have an impact on infant 19 20

I am extremely grateful to A. Nemtsov for graciously providing me with the regional alcohol data. As discussed in section 2, alcohol consumption did recover after the prohibition, but slowly.

mortality (Shkolnikov and Nemtsov 1997), sample bias from surviving into childhood is not a concern in comparing outcomes between the prohibition and non-prohibition cohorts. In addition, since both prior to and during the prohibition, internal migration in Russia was severely restricted, the prohibition was not associated with changes in location in response to alcohol policies, further diminishing concerns about sample selection.21 After the breakup of the Soviet Union in 1990, mobility did increase, however, so comparing child outcomes during and after prohibition could result in biased estimation. I address this issue in several ways. First, I perform estimations on the pre-1990 cohorts only, and show that results are very similar to those using the full sample. Second, since my data contains information on children’s place of birth as well as the household’s current residence, I can also exclude the “movers” from the estimation. The results, however, are not very sensitive to restricting the estimation in these ways. 22 4.2. Identification (II): Regional Variation in Exposure to Prohibition

Since birth cohort variation might simply reflect the effect of macro economic shocks, I use the variation in the intensity of prohibition across regions to difference out time-specific factors that were common across all regions. Although prohibition measures were set at the federal level, their implementation at the local level was uneven, resulting in large variations in the degree to which alcohol consumption decreased after 1985.23 Most importantly for my purposes, the implementation of the prohibition across regions was driven by factors—administrative, monitoring, and political—unrelated to

21 Siberia did experience in-migration during this period, due to both labor needs in the region, as well as political reasons. I therefore experimented with excluding Siberia from the estimation, but the results are essentially unchanged. 22 This is not surprising given that internal migration started increasing significantly only after 1994, a time period which is outside the sample. Even during the early period of transition, entry to cities and certain regions was still restricted; these restrictions were only eliminated during the second half of the 1990s (see Gang and Stuart 2002 for more details). 23 For instance, during the first six months of the campaign, alcohol consumption decreased by 3.6 liters per capita in the Central Black-Earth region, and by only 0.1 liters in the North-Western region. Between 1984 and1986, alcohol consumption had fallen by 5.7 liters per capita in the Volga region, but only by 1.9 liters in the Far Eastern Siberia region.

child health outcomes. To begin with, federal directives and alcohol laws were often vague, and as a result, there was a large variation in their administration at the local level due to often-contradictory interpretations.24 In addition, there were also wide variations in the severity with which the sanctions were applied.25 The reason for this was that not just jurists and the police administered the laws, but also factory managers and trade union officials, who had no legal training, and were often very interested in taking into account the moral aspects of the case (White 1996). As a result, as the justice minister during this time period noted, “every place, town, and even enterprise implement[ed] the legislation in its own way.”26 Monitoring and political considerations also played a role. Larger towns closer to the capital, for instance, were easier to monitor than the countryside and the mountainous regions, and therefore experienced larger changes in alcohol consumption. In addition, local party supervision of alcohol policy often depended on the officials’ personal ambitions, and the degree to which their career goals aligned with the various factions in the central party (White 1996, Tarschys 1993). Finally, the variation in alcohol consumption across regions also depended on the local availability of sugar, which was the most common ingredient for making samogon. Since samogon production was simple and required few skills and equipment, it expanded rapidly, especially during the later years of the prohibition period (White 1996). As moonshiners began competing with housewives, however, sugar supply shortages became common and limited

24

The law, for instance, forbade the drinking of ‘spirits’ in public places, but it did not define spirits. As a result, in some places drinking beer in public areas was allowed because beer was regarded as a “weak” alcoholic drink, whereas in other places even the sale of kefir (a yogurt drink that naturally contains a very small amount of alcohol due to fermentation) was prohibited. 25 A worker for instance, might be dismissed in one enterprise, while another one under similar circumstances would simply be reprimanded in another enterprise. 26 Cited in White 1996,154.

samogon production and consumption, especially in non-sugar producing regions.27 4.3. Regression Framework

Since the variation in the intensity of prohibition across cohorts and regions was exogenous to child outcomes, I can use a differences-in-differences approach to estimate the impact of the anti-alcohol campaign on long run child health. Specifically, I estimate regressions of the following form:

Healthicr = α 0 + β * Highr + γ * Pr ohibitcm + δ ∗ Highr * Pr ohibitcm + + α1 * X icr + α 2 ∗ Rr + α 3 * Cc + ε icr

(1)

In this formulation, i, c, r and m index children, birth year cohorts, regions, and months of birth respectively;28 Health is a measure of the child’s health status (such as immunizations and height), High is an indicator for the intensity of prohibition in a given region (as described in section 3), and Prohibit is an indicator for birth dates between April 1985 and December 1987; the vector Xicr includes time and region varying child-specific factors (such as age and gender), as well as household-specific factors (like income, parental education, and parental height); and the vectors Rr and Cc represent region and cohort fixed effects, respectively. Observations are weighted using survey sample weights, and the standard errors are clustered at the region level. The coefficient of interest in equation 1 is δ, which captures the differential effect of the antialcohol campaign on the outcomes of children born during prohibition in regions where the alcohol consumption dropped most (high intensity regions). In estimations based on equation 1, I control for a rich set of observable parental and household characteristics that are associated with child health outcomes and parental alcohol 27

During the last year of the campaign, sugar sales increased by as much as they had done during the entire decade from 1970 to 1980 (Pravda, September 1988). 28 The dataset contains 2,397 children under the age of 16, and 12 cohorts (birth years 1981-1992). It also covers 8 geographical regions: Metropolitan areas (Moscow and St. Petersburg), Northern and North Western, Central and Central Black Earth, Volga-Vaytski and Volga Basin, North Caucasian, Ural, Western Siberian, Eastern Siberian and Far Eastern. The dataset also covers 160 census regions, and in some robustness check specifications, r will refer to census rather than geographical regions.

consumption. However, unobservable household factors like home environment, parents’ genetic endowment, etc. could still bias the estimation. To address this issue, I also estimate equation 1 on the sample of families that have at least two children, controlling for household fixed effects and using the intra-family variation in child exposure to prohibition to difference out familyspecific (cohort-invariant) effects.

5. Results 5.1. Trends in Regional Alcohol Consumption and Child Outcomes

As shown in sections 1 and 4, the anti-alcohol campaign of 1985-1988 resulted in large changes in alcohol consumption across regions and time. Did these changes result in improved health outcomes for the children who were most exposed to prohibition? Figure 4 provides some suggestive evidence based on differences in height for age (HFA) z-scores between high and low (prohibition intensity) regions, by birth cohort. HFA z-scores were higher (and thus stunting lower) for children born under the prohibition compared to those for either pre- or postprohibition cohorts. Furthermore, whereas trends in the HFA z-scores were essentially identical in the high and low regions for the post-prohibition cohorts, stunting was much lower in the high regions relative to the low ones for the prohibition age groups. Since examining average HFA z-scores can hide a lot of variation in their distribution across cohorts and regions, panels A-C of figure 5 also show the HFA densities across high and low regions, by birth cohort. This comparison shows that there was a dramatic right shift in the entire distribution of HFA z-scores in high intensity regions relative to low intensity ones for children born during the prohibition (relative to children born during other time periods in the sample).

5.2 The Effect of Prohibition on Child Height (I): Basic Results

Figures 4 and 5 thus suggest a strong positive correlation between decreases in the intensity of the prohibition for children born between 1985 and 1988, but not for children born during previous years. Table 3 shows this more formally. The first column presents the OLS results from estimating equation 1, with HFA z-scores as the dependent variable. The vector of controls includes correlates of child height and parental alcohol consumption: child gender, parents’ employment status, indicators for mother’s educational attainment (primary, secondary/vocational, technical, college and above) and marital status, indicators for mother’s height tercile, dummies for family size, and an indicator for whether the child’s birth place was rural or urban. I also include an indicator for whether the mother was less than 20 or over 40 years of age at the time of the child’s birth, since this is known to be correlated with low birthweight and other complications.29 In column 2, I also control for total household income, measured in real terms, and adjusted for differences in the cost of living across regions. To account for the possibility that child health is determined by permanent, rather than current income, in column 3 I use a proxy for the longer-term economic status of the household. This proxy is an asset index, constructed on the basis of the extensive asset information in the data, using principle component methods.30 Overall, the OLS regressions from columns 1-3 suggest that the prohibition had a mildly positive impact on child health in low regions, and a very strong effect in the regions where the alcohol consumption decreases were largest. The results from all columns are pretty similar, and 29

To control for the possibility that fathers and mothers might have differential impacts on child outcomes, I have also performed specifications controlling for father’s educational attainment and height tercile. Results are very similar to those in table 1, but some collinearity problems occur due to the fact that these variables are very highly correlated with their counterparts for mothers—most likely due to assortative matching in Russian marriages. 30 The variables that I use in the principal components analysis are: the size of the house lot , indicators for the availability of heat, cold and hot water, and sewage disposal, and indicators for the ownership of stove, fridge, washing machine, color TV, black and white TV, car, motorcycle, tractor, and second residence or summer home. Filmer and Pritchett (2001) use a similar method using NHFS data, and argue that the asset index might be a bettor measure of household permanent wealth compared to current income and consumption measures.

show that, on average, the prohibition reduced growth stunting relative to the reference median by 0.1 and 0.5 standard deviations in the low and high intensity regions, respectively.31 As discussed in section 2, the positive effect of prohibition on parental time and income resources was partially mitigated by alcohol price increases, queuing to buy alcohol, and especially the replacement of alcohol with toxic substances like glass cleaners and glue. Since higher incomes can partially mitigate these effects, I check whether the effect of the prohibition on child height depended on family (long run) income. As the results in column 4 of table 3 show, income did enhance the effect of prohibition on child height in high areas, but the magnitude of the effect was essentially negligible (0.01 standard deviations at the mean value of the asset index). I will examine the role played by parental income on child outcomes in further detail in section 5.5. To get a better sense of the distribution of the effect of prohibition on child height, I also run two probit estimations of equation 1, with mild and severe stunting as the dependent variables. Mild (severe) stunting occurs when a child’s height is one (two) standard deviations below the reference median for that age. The results, shown in columns 5 and 6 of table 3, show that the prohibition decreased the probability of stunting by 4 percent in low intensity regions, and by over 11 percent in high intensity ones.32 Furthermore, children born during the prohibition in high intensity regions were also almost 1 percent less likely to be severely stunted.33 5.4. The Effect of Prohibition on Child Height (II): Robustness Checks

The Effect on Height, by Intensity of Prohibition Over Time Since the intensity of the prohibition varied not just across regions, but also across birth 31

The mean HFA z-score for prohibition cohorts in low and high intensity regions were –0.35 and –0.23 respectively. The mean probabilities of stunting among prohibition cohorts in high and low regions were 8 and 11 percent. 33 It is not surprising that the magnitude of this effect is smaller, since the occurrence of severe stunting in Russia during this period was also small, averaging about 2.5 percent. 32

cohorts (even during the prohibition period), I check whether this variation translated into differences in child health outcomes.34 Figure 6 provides a graphical depiction of this analysis, by showing the coefficients on the interaction between Prohibit and High for each birth cohort between 1981 and 1992, and their corresponding 95 percent confidence bands. As Figure 6 illustrates, the effect of the prohibition on child height in high intensity areas was only positive for the prohibition cohorts, and it was most strongly statistically significant for the 1986 birth year cohort. This makes it more likely that the effects on child health I find in Table 3 are indeed due to the alcohol campaign, since the cohorts not exposed to prohibition did not experience height improvements. Since the sample sizes are small, however, the results from Figure 6 are rather imprecise. In table 4, therefore, I perform several other robustness checks regarding the effect of prohibition on child height according to the length of exposure. In column 2 of table 4, I show that the effect of prohibition on cohorts born prior to 1985 was not statistically significant. This is not surprising since although these cohorts were exposed to the prohibition for its entire duration, they were also exposed to the negative effects of the very high regimen of alcohol consumption prior to the prohibition.35 In column 3, I re-estimate equation 1 with the variable Prohibit modified to be an indicator for birth dates between April 1985 and December 1986 (rather than December 1987). Although the prohibition officially ended in early 1988, by 1987 the samogon consumption was already increasing, and the severity of alcohol restrictions was decreasing. The results from column 3 show that the effect of the prohibition was indeed stronger among these more exposed prohibition cohorts.36 In addition to varying across cohorts, the degree of exposure to prohibition varied within 34

Column 1 of table 4 simply reproduces the results from column 3 of table 3 to facilitate comparisons. The effect of the anti-alcohol campaign on prohibition cohorts persists even in the presence of an indicator for pre-prohibition cohorts (and its interaction with High). 35

certain birth cohort groups as well. The increases in alcohol prices in August 1985 and August 1986, for instance, which were both associated with a stepping up of the intensity of the campaign, allow me to compare differences among child outcomes within these cohorts. For these purposes, in columns 4 and 5 I estimate equation 1 restricting the sample to children born in 1985, and 1986 respectively. In these specifications, Prohibit becomes an indicator for birth dates prior to August 1985, and prior to August 1986, respectively. The results show that a higher intensity of prohibition was associated with higher within-cohort variation in child outcomes (columns 4 and 5). Furthermore, the first increase in the intensity of the campaign (in 1985) was associated with larger improvements in child health than the second one. Given that the first price increase was compensated, while the second one was not, this provides further evidence that certain prohibition measures actually worked against improving health outcomes. To further exploit the variation in the intensity of prohibition both across and within cohorts, I replace the indicator Prohibit with a continuous measure—the number of days spent under the prohibition regimen (Prohibit Days). Since, as discussed in section 3, the effect of prohibition is potentially largest on children younger than three, I also use as a measure for the intensity of prohibition the number of days between ages zero and three spent under the prohibition (Exp Days). The results from these specifications, presented in columns 6 and 7 of table 4, are consistent with a stronger impact of prohibition for children who were exposed to prohibition longer. Moreover, a longer exposure to prohibition prior to the age of three was particularly beneficial, leading to larger improvements in height.37 Household Fixed Effects, Semiparametric Estimation, and Other Robustness Checks Table 5 presents a number of further robustness checks of the results. In column 1, I estimate equation 1 controlling for household fixed effects. The sample is restricted to families 37

In high regions, the increase in child outcomes at the mean values of the total (between ages 0 and 3) number of days spent under prohibition was 0.13 (0.09) standard deviations respectively.

who have at least two children, and the intra-family variation in child exposure to prohibition allows me to difference out family-specific (cohort-invariant) unobservable factors like the household environment. The results are very similar in spirit to those reported in table 3, but the estimated effects are slightly larger, suggesting that unobservable family factors do play a role in determining the effect of prohibition on child outcomes—albeit a small one. In column 2, I check the sensitivity of my main results to the way I classify regions into high and low intensity regions. In this estimation, the indicator for the intensity of prohibition, High, is replaced with actual regional alcohol consumption in the child’s region of origin, measured during the year of birth.38 In column 3, to account for the possibility that the results are driven by the persistence of alcohol consumption over time, I also control for lagged values of alcohol consumption. The results from columns 2 and 3 are very similar to those in table 3, and indicate that at the mean level of alcohol consumption, prohibition increased the height of children relative to the reference median by 0.55 standard deviations; at higher levels of alcohol consumption (one standard deviation higher), the effect was 0.66. Furthermore, only the alcohol consumption during the year of birth and its first lag had a statistically significant effect on prohibition cohorts. This is consistent with the fact that parental alcohol consumption should affect child outcomes only during early childhood and in utero, but not prior to that. To test the sensitivity of the results to functional form specification, I also estimate equation 1 using the logarithm of height as the dependent variable (rather than z-scores), and controlling for age in a flexible way. Specifically, I use Robinson’s (1988) semiparametric estimator, which allows me enter all controls linearly while leaving the functional form of the relationship between height and age unspecified. The estimates from column 4 are also very 38

I have also performed estimations where regions are classified as “high” if the average drop in alcohol consumption during prohibition was high (above the regional median) relative to 1983 (rather than relative to 1984); if the overall drop by 1986, and by 1987 respectively was above the median; and if the yearly change during prohibition (1985 consumption relative to 1984, and 1986 consumption relative to 1985 etc) was above the median. In all these specifications, results are essentially the same as those using the original classification into high and low intensity regions.

similar in spirit to those in table 1. Prohibition cohorts born in high intensity areas experienced a 5 percent increase in height (4.5 cm), which corresponded to an increase of about 0.5 standard deviations. In column 5, I also test the sensitivity of the main results to the use of alternative regional indicators. Instead of using the eight geographical region indicators discussed in section 4.3, I use census areas fixed effects. This estimation procedure uses a smaller number of observations per area (since I now have 160 rather than 8 regions), but the results are remarkably stable. Finally, in columns 6-8 of table 5, I check whether the results are driven by the impact of post-prohibition factors. As discussed in section 1, due to the hyperinflation in the 1990s that lowered real alcohol prices, alcohol consumption increased rapidly. In addition, economic reforms designed to ensure the transition to a market economy started being implemented in the late 1980s, and especially after the break-up of the Soviet Union in 1990. To address the potential impact of these post-prohibition factors on child outcomes, I use two approaches. First, in columns 6 and 7, I control for two measures of current alcohol consumption: total spending on alcohol, and its share in household budgets. Neither of these measures had an impact on child health, however, nor did their interactions with Prohibit. Second, in column 8, I perform a placebo test by using weight for age (WFA) z-scores as the dependent variable. Unlike HFA z-scores, WFA ones represent a short-run measure of child health, reflecting current flows of health investments, rather than accumulation over time (Falkner and Tanner 1986). As the results in column 8 show, the effect of prohibition on the weight of children born in high intensity regions was not statistically significant, further confirming that the height results are driven by prohibition, and not by the post-1988 changes in household’s economic circumstances and alcohol consumption patterns.

5.3. The Effect of Prohibition on Chronic Health: Results and Robustness Checks

Although the prohibition had a substantial impact on child height, it is important to learn whether it also had an impact on more specific health indicators like immunizations and morbidity. I begin by estimating equation 1 by probit, with the dependent variable being one of two indicators for poor health: whether during the past year the child had any chronic health problems, or whether the child had been hospitalized. I also use as the dependent variable an indicator for the child having had chronic chest congestion, the only measure in my dataset containing information on the occurrence of a specific chronic health condition. The results, shown in columns 1-3 of table 6, indicate that the prohibition had no discernible impact on child chronic health in low intensity areas. By contrast, cohorts born in high intensity regions were 12 percent less likely to report chronic health conditions, 2 percent less likely to have been hospitalized during the previous year, and 6 percent less likely to have suffered from chronic chest congestion. Since self-reported health has been shown in numerous studies to be a good predictor of both short-run and long-run health (see Miilunpalo et al. 1997), I also estimate the effect of prohibition on this alternate health indicator using an ordinal probit specification. The results, shown in column 4, indicate that prohibition increased the likelihood that children born in high intensity areas reported being in good health.39 To confirm that these effects on child health are due to prohibition, I also perform several falsification tests. Specifically, I use as dependent variables acute health outcomes—coughing, ear aches, sore throat and diarrhea—which should not have been affected by past shocks like the prohibition. The results in columns 5-8 of table 6, show that this was indeed the case. Together, the regressions in table 6 suggest strong improvements in child health due to 39

The self-reported health index is defined so that lower values indicate better health.

effects of prohibition. It is important to note here, however, that the cumulative effect of the prohibition over the life of a child could be even larger than the effects (measured in 1995) suggest. The reason for this is that poor health in childhood is associated not just with worse chronic adult health, but also with adverse health trajectories; by middle age, the cumulative impact of childhood shocks on chronic health could be 4-6 times larger than that earlier in life (Haas 2007)

5.4. The Effect of Prohibition on Immunizations

Finally, in column 9 of table 6, I examine the impact of prohibition on the probability of the child being immunized. As discussed in section 3, since vaccinations have an age-specific schedule, they can provide us with a cleaner identification of the effect of prohibition on child health. Specifically, if we find that children who were “eligible” to be vaccinated between 19851988 in high intensity areas had different immunization rates when measured in 1995, we can be confident that this effect was indeed due to the prohibition. The results in column 9 show that this was indeed the case; prohibition cohorts born high intensity areas were 13 percent more likely to have been immunized. The results in column 9 suggest that immunizations can provide a plausible mechanism for the effect of prohibition on height and child chronic conditions that I found in sections 5.1 to 5.3. The reason for this is that increased immunization rates are associated with declines in childhood infectious diseases, which, in turn, are associated with increases in growth, and decreases in chronic conditions later in life (Martorell et al. 1994, Blackwell et al. 2001, Barker 1995, Costa 2000). 40

40

Frequent and severe infections during early childhood have been shown to impair growth (Martorell et al. 1994), since they can both lead to and exacerbate inadequate dietary intake (Scrimshaw et al. 1968). In addition, childhood infections are also

5.5. The Effect of Prohibition on Child Health: Time or Money?

The results in the previous sections suggest that the prohibition had a large impact on child health in the regions that experienced large drops in alcohol consumption, possibly due to its effect on immunizations. What were the channels through which this effect took place? As discussed in section 2, parental alcohol consumptions reductions can affect child outcomes by increasing parental time and monetary investments in child health. The question that naturally arises is which of these two factors played a larger role in improving child health during the prohibition: was it time or money? Although I cannot provide a definite answer to this question due to data limitations, I can nevertheless provide some very suggestive evidence based on the effect of prohibition on immunizations. In the 1950s and 60s, the health care system in the Soviet Union had been particularly successful in reducing infant deaths from infectious diseases through a variety of public health and mandatory mass immunization campaigns (Brainerd and Cutler 2005). By the late 1970s, however, the worsening of the economic conditions in Russia started putting a strain on the resources available for immunizations (Vitek and Wharton 1998).41 The result was an increase in rationing for some vaccines, such as polio, that had been mandatory in previous decades. Furthermore, some newer vaccines (for measles and hepatitis for instance), which had not been part of the general immunization campaigns in the 1950s-1960s, were available only on an “optional” basis, and usually required an informal (bribe) payment.42 Importantly for my purposes, due to the centralized nature of the healthcare provision system, these changes in the provision of vaccines were determined by federal (rather than local) factors, and their effect was thus uniform across the Soviet Union. associated with chronic conditions later in life such as heart diseases, cancer and lung conditions (Blackwell et al. 2001, Barker 1995, Costa 2000). 41 In fact, the outbreak of a diphteria epidemic in Russia during the 1990s, for instance, was largely attributed to decreased immunizations in the 1980s (Vitek and Wharton 2005). 42 Vitek and Wharton 1998, Ryan 1998.

In table 7, I take advantage of this difference in time and monetary costs among vaccines to learn more about the channels through which prohibition affected child health. The results in columns 1-3 show that prohibition had a strong impact on polio immunization, a process that was time-intensive due to queuing. By contrast, vaccination outcomes that were more intensive in parents’ money—measles and hepatitis—either decreased, or were unaffected by prohibition. To check for the possibility that government, rather than parental effort, drove the immunization results, I use additional information on polio vaccinations to perform two specification checks. In column 4 of table 7, I restrict the sample to children who reported not having been immunized at school, since vaccination outcomes for these children were more likely to be driven by government (rather than parental) action. In column 5, I also check to see whether the effect of prohibition on immunization outcomes was higher for families with more children, since these families had an increased chance of governmental health intervention (and immunization) through social worker involvement. The results from columns 4 and 5, however, are very similar to those in column 1. Overall, the results in table 7 provide suggestive evidence that time factors might have played a larger role in improving child health compared to parental income, at least where immunizations were concerned.

6. Conclusion The main contribution of this paper is to provide new evidence that parental alcohol consumption during very early childhood can have long-term effects on a child’s physical health, even in very heavy drinking environments. My paper takes advantage of a unique temporal shock to alcohol supply in Russia (the 1985-1998 prohibition), which featured large exogenous variations in the efficacy of the prohibition program across regions. This enables me to eliminate the endogeneity

problems that are usually central to these sorts of studies. In addition, I provide evidence from vaccination rates that alcohol negatively affected parental time investments in children more than resource investments, which is a subject of great interest for further research, ideally involving panel data at the household level.

References: Barker, DJP (1995). "Fetal Origins of Coronary Heart Disease." British Medical Journal 311(6998): 171-174. Blackwell, D. L., M. D. Hayward, E. M. Crimmins (2001) “Does childhood health affect chronic morbidity in later life?” Social Science & Medicine 52 (2001) 1269-1284. Beard, Albertine S. and Martin J. Blaser (2002). “The Ecology of Height: The Effect of Microbial Transmission on Human Height.” Perspectives in Biology and Medicine 45 (Autumn): 475-99. Bonu S, M. Rani, P.Jha, D. H. Peters, N.S. Nguyen (2004). “Household Tobacco and Alcohol Use, and Child Health: an exploratory Study from India”, Health Policy, 70, 67-83. Brainerd, Elizabeth; David M. Cutler (2005). "Autopsy on an Empire: Understanding Mortality in Russia and the Former Soviet Union," Journal of Economic Perspectives, vol. 19(1), pages 107-130, Case, Anne and Christina Paxson (2006) “Stature and Status: Height, Ability, and Labor Market Outcomes”, Princeton mimeo. Chatterji, Pinka and Sara Markowitz (2001) “The Impact of Maternal Alcohol and Marijuana Use on Children's Behavioral Problems, ” Journal of Health Economics, 20, No. 5, 703-731. Costa, Dora, and Joanna Lahey (2005) “Becoming Oldest Old: Evidence from Historical US Data” Genus. 2005. 51(1): 125-61. Costa, D. L. (2000). “Understanding the twentieth century decline in chronic conditions among older men”. Demography, 37, 53-72. Doblhammer G., Vaupel J.W. (2001), “Lifespan depends on month of birth”, Proceedings of the National Academy of Sciences, 98(5), 2934-2939. Ducham, N.Y., and F.E. Sheregi (1986). “Prichini i sotsial'niye posledsviya piyanstva” [Causes and social consequences of drunkenness . Sotsiologicheskiye issledovaniya [Sociological Studies] 2:144-152. Eurocare (1998).. Alcohol Problems in the Family A Report to the European Union. England Falkner, Frank and J. Tanner (1986). Human Growth: A Comprehensive Treatise, vol.3, NY: Plenum. Filmer, Deon, and Lant Pritchett (2001). “Estimating Wealth Effects without Expenditure Data or Tears: An Application to Educational Enrollments in States of India”, Demography, Vol. 38, No. 1, pp. 115-132 Floud, R., K. Wachter, and A. Gregory (1990). Height, Health and History. Cambridge Studies in Population, Economy, and Society in Past Time. Nutritional Studies in the United Kingdom 1750–1980. Cambridge, England: Cambridge University Press.

Fogel, R.W., and D.L. Costa (1997). “A Theory of Technophysio Evolution, With Some Implications for Forecasting Population, Health Care Costs, and Pension Costs.” Demography 34:49–66. Gang, Ira and Robert C. Stuart (2002) “The Political Economy of Russian City Growth”, Economic Development and Cultural Change 50, 491-508. Grossman, Michael and Sara Markowitz (1998) “Alcohol Regulation and Domestic Violence Towards Children,” , Contemporary Economic Policy, 16, No. 3, 309-320. Goskomstat Rossii (1992) ,Pokazateli Sotsial'nogo Razvitiya Respublik, Krayev, I Oblastey Rossiyskoy Federatsii. Moscow Goskomstat Rossii (1993) ,Pokazateli Sotsialinogo Razvitiya Rosskiyskoy Federatsii I Ee Regionov., Moscow Goskomstat Rossii (1995), Rossiyskiy Statisticheskiy Ezhegodnik, Moscow Grossman, Michael and Sara Markowitz (2000) “The Effects of Beer Taxes on Physical Child Abuse,” (with Michael Grossman), Journal of Health Economics, 19, No. 2 (March 2000), 271-282. Haas, Steven A (2007) “The Long-Term Effects of Poor Childhood Health: An Assessment and Application of Retrospective Reports”, Demography , Volume 44, Number 1, February, pp. 113135

Huang, LX, Cerbone, FG, Gfroerer, JC (1996). Children at Risk Because of Parental Substance Abuse. OAS Working Paper, Substance Abuse and Mental Health Services Administration. Ivanets, L.M and M.I. Lukomskaya (1990) “The USSR’s New Alcohol Policy.” World Health Forum, 11 , 246-52 Jones, A.S., D.J. Miller, D.S. Salkever (1999) “Parental Use of Alcohol and Children's Behavioral Health: A Household Production Analysis”. Health Economics, 8:661-83. Kuh, D.J. and M.E. Wadsworth (1993). "Physical Health Status at 36 Years in a British National Birth Cohort." Social Science and Medicine 37(7):905-916. Markowitz, Sara (2000). “The Price of Alcohol, Wife Abuse and Husband Abuse”, Southern Economic Journal, 67, No. 2 (October),279-303. Martorell, R., L. Kettel Khan and D.G. Schroeder (1994). “Reversibility of Stunting: Epidemiological Findings in Children from Developing Countries.” European Journal of Clinical Nutrition 48 Manton, K.G., E. Stallard, and L. Corder (1997). “Changes in the Age Dependence of Mortality and Disability: Cohort and Other Determinants.” Demography 34:135–57. Miilunpalo, S; I Vuori; P.Oja, M. Pasanen and H. Urponen (1997) “Self-rated health status as a health measure” Journal of Clinical Epidemiology, Volume 50, Issue 5, Pages 517-528

Nemtsov, A (1992). “Uroven realnogo potrebleniya alkogol'ya v Rossiysoy Ferartsii”. Sotsialnaya I klinicheskaya pychiiatriya 2, 46–53 Nemtsov, A., and A. Nechaev (1991) “Alkogolnaya situatsiya v Moskve v 1983-1990 godah”. Sotsialnaya i klinicheskaya psychiiatriya [Social and Clinical Psychiatry] 1:75-83 Nemtsov, Alexander; Vladimir M. Shkolnikov (1997) “The Anti-Alcohol Campaign and Variations in Russian Mortality,” in José Luis Bobadilla, Christine A. Costello, and Faith Mitchell, editors, Premature Death in the New Independent States. National Academy Press, Washington, D.C, pp.239-261 Ravelli, ACJ and JHP van der Meulen, RPJ Michels, C. Ostmond, DJP Marker, CN Haels and OP Bleker. 1998. "Glucose Tolerance in Adults After Prenatal Exposure to Famine." The Lancet 351 Richardson, Erica (1999). “The Struggle For Sobriety: Anti-Alcohol Campaigning Under Gorbachev And Yeltsin”. Research Papers In Russian And East European Studies, March. Robinson, P. M. (1988) “Root-N-Consistent semiparametric regression”, Econometrica, Vol. 56, No. 4. Scrimshaw, N. S., Taylor, C. E., & Gordon, J. E. (1968). Interactions of Nutrition and Infection. Geneva: World Health Organization. Segal, B. M (1990) The Drunken Society. Alcohol Abuse and Alcoholism in the Soviet Union. New York : Hippocrene books. Substance Abuse and Mental Health Services Administration (SAMHSA), Office of Applied Studies (2004), “Alcohol Dependence or Abuse among Parents with Children Living in the Home.” National Survey on Drug Use and Health (NSDUH) Report, February Tarschys, Daniel (1993) “The Success of a Failure: Gorbachev's Alcohol Policy, 1985-88” Europe-Asia Studies, Vol. 45, No. 1, pp. 7-25. Treml, Vladimir G (1997). “Soviet and Russian Statistics on Alcohol Consumption and Abuse,” in José Luis Bobadilla, Christine A. Costello, and Faith Mitchell, editors, Premature Death in the New Independent States. National Academy Press, Washington, D.C, 1997, pp.220-238. Treml, Vladimir G (1982). Alcohol in the USSR: A Statistical Study, Durham, N.C.: Duke University Press, 1982. Vitek CR, Wharton M. 1998. “Diphtheria in the former Soviet Union: reemergence of a pandemic disease” Emerging Infectious Disease Oct-Dec;4(4):539-50. White, S (1996). Russia Goes Dry. Alcohol, State and Society. Cambridge University Press, Cambridge. Zaigrayev, G.G (1992). “Obshestvo I Alkogol” [Society and Alcohol]. NIIMVD [Research Institute of the Ministry of Interior Affairs], Moscow. Zohoori, N., L. Henderson, K. Gleiter, and B.M. Popkin (1999). "Monitoring Health Conditions in the Russian Federation: The Russia Longitudinal Monitoring Survey 1992-98." Report submitted to the U.S. Agency for International Development. Carolina Population Center, University of North Carolina at Chapel Hill, North Carolina

4

Alcohol Consumption per Capita (liters of 100% ethanol) 6 8

10

Figure 1. Official Alcohol Consumption per capita in Russia, 1960-1995

1960

1965

1970

1975 1980 year

1985

1990

1995

Note: Alcohol consumption data was published by Goskomstat Rossii (1992, 1993, 1995) and Treml (1982), and reproduced in Treml (1997). Alcohol data is measured in liters, and was derived from sales of all state-produced alcoholic beverages, i.e., vodka, fruit wine, grape wine, cognac, champagne, and beer, converted to 100% alcohol.

30

0

Consumption per capita (liters) 5 10 15 20

25

Figure 2 Per Capita Consumption of State-Produced Alcoholic Beverages in Russia, 1970-1995

1970

1975

1980 beer wine

year

1985

1990

1995

vodka

Note: The data was published by Goskomstat Rossii (1992, 1993, 1995) and Treml(1982), and reproduced in Treml (1997). Homedistilled samogon and home-made wine are excluded. Per capita consumption of fruit wine, cognac, and champagne is not shown separately.

31

4

Consumption per capita (liters of 100% alcohol) 6 8 10 12

14

Figure 3 Official and Actual Per Capita Alcohol Consumption in Russia, 1960-1995

1970

1975

1980

year

1985

1990

1995

Alcohol and samogon consumption Adjusted alcohol consumption Official alcohol consumption Samogon consumption

Note: Official alcohol consumption data was published by Goskomstat Rossii (1992, 1993, 1995) and Treml(1982), and reproduced in Treml (1997). It was derived from sales of all state-produced alcoholic beverages, i.e., vodka, fruit wine, grape wine, cognac, champagne, and beer, converted to 100% alcohol. Alcohol and samogon consumption is the sum of official alcohol consumption and samogon consumption. The samogon data is from Goskomstat estimates and was reproduced by Treml (1997). Samogon consumption refers to sugar-based samogon only. The estimates exclude samogon produced from other inputs, such as potatoes, grain, and fruits, as well as home-made wines and beers. Adjusted alcohol consumption is total alcohol consumption (samogon included) as estimated by Nemtsov(1992, 1997), and Nemtsov and Nechaev (1991).

32

-.6

Height For Age Z-score -.4 -.2 0

.2

Figure 4. Height for Age Z-scores, by Intensity of Prohibition across Regions

1982

1984

1986 Year of Birth

1988

1990

High Intensity Regions Low Intensity Regions Note: The figures shows average height for age (HFA) z-scores for children born in high and low prohibition regions. Data on height and age is from round 5 of the RLMS. The HFA z-score represents the number of standard deviations from the NCHS reference median height for a given age, as suggested by the World Health Organization. The alcohol consumption data is from Nemtsov(1992, 1997), and Nemtsov and Nechaev (1991) and is used to classify regions by intensity of prohibition. High (low) is an indicator for regions in which the change in alcohol consumption during the prohibition (relative to 1984) was above (below) the median change across regions.

33

Figure 5. HFA Z-scores in High and Low intensity Regions, by Birth Cohort Panel B: Panel A: Children Born between 1985-1987

0

0

Density .2

Density .2

.4

.4

Panel A: Children Born between 1988-1992

-2

0

2

-2

High intensity Regions Low intensity Regions

0

2

High intensity Regions Low intensity Regions

0

Density .2

.4

Panel C: Children Born between 1982-1984

-2

0

2

High intensity Regions Low intensity Regions

Note: The figure shows kernel density estimates for (HFA) z-scores for children born in high and low prohibition regions, by cohort of birth. Data on height and age is from round 5 of the RLMS survey. The HFA z-score represents the number of standard deviations from the NCHS reference median height for a given age, as suggested by the World Health Organization. The alcohol consumption data is from Nemtsov(1992, 1997), and Nemtsov and Nechaev (1991) and is used to classify regions by intensity of prohibition. High (low) is an indicator for regions in which the change in alcohol consumption during the prohibition (relative to 1984) was above (below) the median change across regions.

34

Figure 6. Differential Effect of Prohibition in High Regions, by Year of Birth

Coefficient on Interaction Birth Year *High Region

1.5 1 0.5 0 -0.5 -1 -1.5 1979

1981

1983

1985

1987

1989

1991

Year of Birth Note. Figure 6 graphs the coefficients on the interaction between birth year and High. The regression framework is similar to that in table 3, with the addition of the interactions as additional controls. High is an indicator for regions in which the change in alcohol consumption during the prohibition (relative to 1984) was above the median change across regions. Observations are weighed using survey sample weights. Dotted lines depict the 95 percent confidence bands.

35

Table 1. Summary Statistics (I): Household and Regional Alcohol Consumption Variable

Obs

Mean Std. Dev.

Controls

1= residence in urban area 1=mother’s education: primary 1= mother’s education: secondary /vocational 1= mother’s education: technical 1= mother’s education: college and above Number of children Asset index Ln (total hh income) Gender Nb of males in the hh who are employed Nb of females in the hh who are employed 1=mother’s age 40 1=1st tercile of mother’s height 1=2nd tercile of mother’s height 1=3rd tercile of mother’s height Ln(alcohol spending) % of alcohol spending in hh budget

2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379

0.63 0.10 0.24 0.42 0.25 1.82 1.76 8.85 1.49 0.98 1.10 0.12 0.40 0.35 0.26 2.42 0.58

0.48 0.30 0.43 0.49 0.43 0.89 0.92 0.90 0.50 0.49 0.41 0.33 0.49 0.48 0.44 3.11 0.79

2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379 2379

14.61 14.42 14.38 14.30 13.47 10.64 10.74 11.26 11.75 12.10 12.65 13.52

0.89 1.06 1.13 1.25 1.34 0.72 0.76 0.93 0.91 1.03 1.49 1.93

Alcohol Consumption in Year: 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992

Note: data on households controls is from round 5 of the RLMS. Monetary values (income and alcohol spending are expressed in 1995 rubles, and corrected for differences in the cost of living across states. The asset index is an indicator of the long run economic households and was constructed using principal components methods from the following variables: the size of the house lot , indicators for the availability of heat, cold and hot water, and sewage disposal, and indicators for the ownership of stove, fridge, washing machine, color TV, black and white TV, car, motorcycle, tractor, and second residence or summer home. The alcohol consumption data is from Nemtsov(1992, 1997), and Nemtsov and Nechaev (1991) and refers to total alcohol consumption

(samogon included) per capita in the indicated years (expressed in liters).

36

Table 2. Summary Statistics (II): Health Measures

Variable

Height (cm) Weight (kg) HFA z-score WFA z-score 1= chronic health probla. 1=hospitalized a 1=coughingb 1=chronic chest congestionb 1=ear acheb 1=sore throatb 1=diarrheab Self reported health 1=immunized 1=measles immunization 1=polio immunzation 1=hepatitis immunization

Obs

676 676 676 676 676 676 676 676 676 676 676 676 676 676 676 676

(1)

(2) Birth Cohorts

(3)

1981-1984

1985-1988

1989-1992

Mean

150.82 42.64 -0.36 -0.35 0.35 0.03 0.26 0.34 0.03 0.18 0.02 2.52 0.99 0.61 0.64 0.59

Std. Dev.

13.36 11.70 1.20 1.05 0.48 0.18 0.44 0.47 0.17 0.38 0.15 0.63 0.12 0.49 0.48 0.49

Obs

674 674 674 674 674 674 674 674 674 674 674 674 674 674 674 674

Mean

127.46 26.36 -0.29 -0.21 0.39 0.03 0.29 0.33 0.03 0.17 0.03 2.48 0.98 0.73 0.76 0.58

Std. Dev.

9.01 5.16 1.46 1.01 0.49 0.17 0.45 0.47 0.16 0.38 0.18 0.63 0.15 0.45 0.43 0.49

Obs

1029 1029 1029 1029 1029 1029 1029 1029 1029 1029 1029 1029 1029 1029 1029 1029

Mean

97.12 16.03 -0.37 0.00 0.43 0.05 0.31 0.36 0.02 0.14 0.04 2.38 0.97 0.55 0.56 0.37

Std. Dev.

18.89 5.59 1.98 1.44 0.49 0.23 0.46 0.48 0.13 0.35 0.20 0.62 0.17 0.50 0.50 0.48

Note: . a during the past year; b during the past three months. The data on health variables is from round 5 of the RLMS. Columns 1-3 present summary statistics for children born during the indicated years. HFA (WFA) z-scores represent standard deviations from the NCHS reference median height (weight) for a given age, as suggested by the World Health Organization

37

Table 3. The effect of Prohibition on Child Height: Main Results (1) Dependent variable Prohibit High Prohibit * High

(2)

(3)

HFA z-score HFA z-score HFA z-score HFA z-score 0.101*** (0.02) -0.050*** (0.006) 0.38** (0.12)

0.09*** (0.019) -0.049*** (0.006) 0.40** (0.12)

-0.041*** (0.014) 0.0089*** (0.0014) -0.12*** (0.0076) -0.0162 (0.010)

0.009 (0.020) 0.0016 (0.006) -0.007*** (0.003) 0.0005 (0.002)

0.002 (0.094) 0.045 (0.14) 0.58*** (0.10) 0.76*** (0.15) 0.09 (0.13) 0.108 (0.14)

0.095* (0.042) 0.008 (0.094) 0.036 (0.15) 0.58*** (0.10) 0.75*** (0.14) 0.123 (0.13) 0.108 (0.15)

0.013 (0.098) -0.067 (0.17) 0.60*** (0.11) 0.78*** (0.15) 0.141 (0.13) 0.145 (0.13)

0.005 (0.098) -0.06 (0.17) 0.601*** (0.11) 0.78*** (0.15) 0.147 (0.13) 0.149 (0.13)

0.0004 (0.010) 0.0203 (0.022) -0.039** (0.019) -0.044*** (0.01) -0.013 (0.022) -0.042** (0.018)

0.0002 (0.002) 0.003 (0.005) -0.005 (0.004) -0.004 (0.003) -0.001 (0.005) -0.004 (0.004)

0.038 (0.083) 0.34** (0.098) 0.047 (0.088)

0.023 (0.079) 0.31** (0.10) 0.027 (0.083)

0.073 (0.084) 0.36*** (0.09) 0.049 (0.08)

0.084 (0.083) 0.36*** (0.092) 0.059 (0.083)

-0.011 (0.018) -0.034** (0.014) -0.0142 (0.016)

-0.0007 (0.003) -0.006 (0.004) -0.001 (0.004)

2379 0.61

2379 0.63

2379 0.63

2379 0.63

2379 0.71

2379 0.72

Ln (income)

1=height Q2m 1=height Q3m 1=primary educationm 1=secondary/vocational educm 1=technical educationm 1=college + education m 1=age 40m Observations Adjusted/Pseudo R-squared

(6) 1=Severe stunting

0.18*** (0.019) -0.046*** (0.006) 0.34*** (0.091) 0.0508 (0.062) 0.007** (0.002)

Asset Index * Prohibit

1=Urban

(5) 1=Mild stunting

0.12*** (0.021) -0.046*** (0.006) 0.37** (0.12) 0.091 (0.054)

Asset Index

1=Male

(4)

Note. m refers to mother’s characteristics.* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors (in parentheses) are clustered at the region level. The sample in all regressions covers 8 regions and 13 birth cohorts (1980-1992) The dependent variable in columns 1-4 is HFA z-score, i.e. the number of standard deviations from the NCHS reference median height for a given age, as suggested by the World Health Organization. The dependent variable in columns 5-6 are indicators for mild stunting (HFA z-score