International Trade and Risk Sharing in the Global Rice Market: The Impact of Foreign and Domestic Supply Shocks

In the first decade of this millennium, rising food prices returned as a concern for policy makers, especially in developing economies. This paper examines how supply shocks, both domestic and foreign, impacted imports and consumption in the world rice market between 1960 and 2010. Such an investigation is important in assessing the role of trade in compensating for domestic shocks. If shortages lead economies to impose trade restrictions, then trade may not be allowed to play an important role in stabilizing consumption. The existing literature has highlighted the importance of these policy shocks in the world rice market and how they have worked to increase the volatility of prices and trade flows. Although trade cannot be expected to play a strong role when the major producing and consuming economies are simultaneously hit by negative yield shocks, such a scenario has occurred in only about 3% of all observed cases. We also find that consumption fails to stabilize even when domestic shocks are negative and foreign shocks are positive; however, imports do peak. Thus, while trade does help in coping with domestic risks, it is unable to achieve full risk sharing. Therefore, no matter the nature of foreign shocks, the principal concern is to stabilize consumption when an economy is hit by negative domestic yield shocks. The frequency of such shocks is about 12% in all observed cases, highlighting the importance of domestic responses. We find that domestic rice stocks have been important in stabilizing consumption. The reliance on domestic policies has, in turn, kept the rice market thin.


I. INTRODUCTION
In recent years, an old concern has resurfaced-that of rising food prices. After the food crisis in the mid-1970s, the world enjoyed declining to stable real prices until the mid-1990s. In 1995mid-1990s. In -1996, there was a spike in prices followed by a return to the long-term trend. From the early part of the 2000s, however, prices have crept upwards, culminating in sharp rises during 2006-2007to 2008-2009. Palm oil, rice, and wheat prices doubled in 2007-2008relative to 1999-2000. Wheat and maize prices increased by more than 75% (Gilbert 2011). 1 What was striking was that the price spikes happened in a very short time interval. In nominal terms, world maize prices increased by 54% from August 2006 to February 2007 followed by an increase in world wheat prices of 125% from May 2007 to March 2008. The most dramatic increase occurred in rice prices. From April 2001 to September 2007, the gradual upward drift saw the price of Thai 100% B rice doubled from $170 per ton to $335 per ton, amounting to a 67% increase relative to the United States Consumer Price Index. But between October 2007 and April 2008, the price tripled to over $1,000 per ton (Dawe and Slayton 2011).
The food price spikes of 2007-2008 have renewed old debates about the efficacy and desirability of price stabilization measures. Economists have long argued that storage-based price stabilization is expensive and, in some instances, ineffective. On the other hand, opening up the economy to trade can be effective in insulating against severe domestic shocks. The food price crisis of 2007-2008, however, planted doubts in policy makers about the reliability of world markets in times of need. Several policy studies have concluded that some public grain reserves are necessary. Price stabilization pursued through public stocks cannot be effective, however, when borders are open. So some restriction of trade would also be necessary.
In the context of this debate, the goal of this paper is to examine how supply shocks, both domestic and foreign, have mattered to imports and consumption over the period 1960-2010 in the global rice market. In autarkic economies, domestic supply shocks drive consumption shocks as well. In countries open to trade, and when trade functions well, domestic consumption depends on both domestic and foreign supply shocks. In particular, compared to autarky, domestic shocks would matter less because of access to world markets. For small open economies, domestic shocks should not matter at all.
These ideal outcomes may not be obtained, however, if policies impede trade. Rising prices often provoke governments to put in place policies that buffer the impact. When they take the form of trade restrictions, world trade may shrink; thus, countries might not have access to world supplies to compensate for adverse domestic shocks. Rice is commonly considered the archetype of an agricultural staple that is subject to such endogenous policy shocks. Hence, we chose to study the impact of domestic and foreign supply shocks on rice imports and consumption.
The outline of this paper is as follows. The next two sections offer a selective survey of the literature on the global rice market with respect to endogenous policy shocks and the reliability of rice trade. Section IV is a descriptive account of the global rice trade and the trade interventions of major exporters. Section V offers a statistical analysis of the impact of exogenous domestic and foreign supply shocks on imports and consumption. Section VI extends this to include the policy variable of domestic and foreign stocks. Concluding remarks are gathered in Section VII.

II. THE RICE MARKET AND ENDOGENOUS SHOCKS
The role of policy responses in provoking and exaggerating price spikes has been particularly highlighted by the global rice market. A review of the literature tells us that it is the rice market that is particularly subject to endogenous policy shocks. Unlike wheat and maize, a relatively small proportion of world rice production (7%) enters international trade. Moreover, wheat and maize trade is driven by surpluses from rich and large land-abundant countries such as Argentina, Australia, Canada, and the United States (US). In the case of wheat, Australia, Canada, and the US export more than 50% of their production. The biggest rice exporter, Thailand, exports close to 40% of its output. However, its share in world rice output is less than 5%.
On the other hand, the large rice-producing countries such as Bangladesh, the People's Republic of China (PRC), India, and Indonesia are either deficient, or at best, have small surpluses relative to consumption. All of these countries have poor populations that are severely affected when rice prices rise. Due to such food security concerns, these countries will likely reduce their net supply to the world markets in times of crisis. This can take the form of export restrictions for exporters, or reductions in import tariffs. In either case, the attempts of these countries to increase their share of world consumption raise world prices. Thus, policies directed toward insulating domestic markets magnify international price volatility when all countries attempt to insulate their respective domestic markets at the same time (Abbot 2011;Martin and Anderson 2011).
During the crisis of 2007-2008, many scholars argued that it was likely that the spike in rice prices was due not to crop failure or low stocks but to policy measures put in place by panicked governments. Writing as early as October 2008, Timmer (2008) argued that the underlying causes for the rise in rice prices are different from those in wheat and maize prices. Low stocks, crop failure, or financial speculation were not plausible factors behind the price increases in rice in [2007][2008]. Nor could these increases be attributed in a straightforward manner to the rise in wheat or maize prices because substitution in consumption among these grains is limited. Rather, the spike must be seen as due to export restrictions by some of the major exporting countries, which induced panic buying by importers such as the Philippines; and storage-driven because of the hoarding instincts of governments and other agents. This has been echoed by others (Dawe and Slayton 2011;Gilbert and Morgan 2010;Wright 2011). Martin and Anderson (2011) estimate that more than 45% of the explained change in international rice price during 2005-2008 was due to export restrictions (compared to 29% for wheat). If anything, this estimate is surprising in that endogenous shocks account for only about half of the rice price increase when most of the literature seems to argue that it is significantly driven by policy shocks. The hypothesis that export policies contribute to global price volatility has also been tested by Giordani, Rocha, and Ruta (2012). Using a data set on trade measures relating to the food sector, they find that the probability that a country imposes a new export restriction is positively associated with the global restrictions on the product (i.e. the share of international trade covered by export restrictions). Furthermore, for 2008-2010, they estimate that a 1% surge in the share of trade covered by export restrictions is associated with a 1.1% increase in international food prices.

III. THE RELIABILITY OF RICE TRADE AND MARKETS
As mentioned earlier, in an integrated global market, trade provides a means for price stabilization without costly investment in commodity stocks. This has been the view of many economists. However, this does not take into account the possibility of government interventions such as market-insulating policies. If exporters restrict their supply fearing a shortfall, importers are deprived of food just when they need it the most. Such an experience may well persuade importers that food trade is unreliable and that they should invest in domestic stocks. Gilbert (2011) argues that it is the rice market and rice trade that are unreliable among those of the major grains. In an earlier work (Gilbert 2010), he showed that a commonly quoted world rice price-the spot price in Bangkok-follows various national prices rather than the other way around (as it is for maize). As it is the rice market that "functions least well," Gilbert (2011) argues for a pragmatic approach where it is recognized that low-income countries "can probably rely on being able to import additional maize or wheat if this proves necessary, but may justifiably be worried about being able to do so for rice." He argues, "[T]his points towards the need for contingency arrangements for rice-either food security stocks, or formal trade agreements with rice exporters or, where this is feasible, a move towards rice self-sufficiency." A related point is that the rice market has been seen to be somewhat disconnected from the markets for other cereals. Shocks to rice supply and demand are not highly correlated with those to other grains. Global futures markets are irrelevant to rice and the crop does not have a use as a biofuel (Dawe and Slayton 2011). It is in this sense that Gilbert and Morgan (2010) regard the rice price spike in 2007-2008 as "peculiar and in some sense pre-modern." Unlike that of other grains, the price volatility in this market does not always depend on the fundamentals of demand and supply shocks and price elasticities. The particular problem of the rice market is the tendency of important trading countries to shield themselves from external shocks. Hence, "rice is different" and the future course of volatility will depend on how the international community addresses the particular problems of this market (Gilbert and Morgan 2011).

IV. GLOBAL RICE TRADE
Imagine a two-country trade model where one of the countries is producing rice. Imagine also that there is no government intervention in either exports or imports. The production of rice is subject to stochastic yield shocks. It is expected then that the higher the yield, the greater the volume of rice that is traded. Figure 1 plots the proportion of world output that is exported against world yields for 1960-2011. The world yield is the production share weighted average of individual country yields. For world yields up to 3 tons per hectare (ha), world exports fluctuate at around 4% of world output without any trend. Beyond that, in the range of 3 to 3.5 tons per ha, the ratio of exports to world output fluctuates at around a higher level of 7%. A closer look shows that the observations in the right half of the graph, involving world yields of more than 3 tons per ha, belong to the period beginning 1994. 1961 1962 1963 1960 1964 1965 1966 1968 1967 1970 1969 1971 1973 1972 1974 1976 1975 1977 1978 1979 1980 1981 1982 19831984 1986 1987 1989 1985 1988 1990 1992 1993  Table 1 shows that the average export-output ratio in 1994-2011 was 7.16%-which represents an increase of 87% over the average value in the pre-1994 period. The discrete jump in the export-output ratio is primarily due to increased rice exports from India. Up to the early 1990s, quantitative restrictions clamped down on non-basmati rice exports from India. The removal of these restrictions in 1993 and 1994 led to non-basmati rice exports of 4.5 million tons from less than a million tons in the early 1990s (Kubo 2011). The other factor behind the higher export-output ratio is the rise of Viet Nam as a major rice exporter. This has been a more gradual process starting from the country's reentry into the world market in 1989. Export liberalization in India and Viet Nam (the leading exporters next to Thailand), therefore, explains why the world rice market grew relatively "thicker" in the 1990s. However, from Table 1, note that the pre-1994 period is characterized by low variability in the export-output ratio even as yields doubled, while the post-1994 period is characterized by high variability in the export-output ratio even as yields have remained in a narrow range of 3-3.5 tons per ha. The coefficient of variation of the export-output ratio in 1994-2011 is twice that in the pre-1994 period. Thus, it seems that while world markets have been more open since the 1990s, policy interventions have made them more unstable as well. India and Viet Nam were among the first countries to impose export restrictions in 2007. More generally, both these countries have domestic concerns that spill over into international markets. This was evident even prior to the 2007 crisis.

Figure 1: World Rice Trade and World Yields
In India, the principal domestic policy imperative is for the government to procure enough supplies to maintain its distribution channel of subsidized rice and wheat. A failure to restrict procurement left the country with an accumulation of massive stocks. In April 2001, this amounted to 51 million tons of grain, including 25 million tons of rice. The subsequent unloading of stocks in the international market led to rising exports and the prolonged stagnation of rice prices in the global market (Kubo 2011). Such large-scale dumping of government stocks on the world market ceased after 2004. By 2005, rice stocks in India had fallen to 13 million tons and more significantly, wheat stocks had dropped to 2 million tons. A subsequent shortfall in wheat procurement that coincided with wheat crop failures in the rest of the world panicked the government into wheat imports and a determination not to allow similar shortfalls in rice procurement. So after dumping rice stocks into the world market in the early 2000s, the government moved to restrict and finally ban rice exports in the late 2000s. With the recovery of rice and wheat stocks, the government once again lifted export restrictions.
Viet Nam has always maintained tight control over rice exports. Initially this took the form of export quotas for registered companies. These were later abolished, and now the government suspends rice exports once the total reaches the targeted level. In 2007, this happened routinely according to the export target of that year. In 2008, faced with rising domestic prices, the government did not allow new export contracts until July of that year. As in India, concern over the domestic availability of rice prompts the government to tightly monitor export volumes. However, there is a difference as well: India's exports are less than 5% of its consumption, while for Viet Nam, they amount to more than 30% of the country's consumption. Global sales are more important for Viet Nam-correspondingly, their regulation has been more predictable and more sensitive to the interests of exporters.

V. THE IMPACT OF EXOGENOUS SHOCKS ON IMPORTS AND CONSUMPTION
A systematic relationship between world yields and global rice trade is not evident in Figure 1. Within a two-country model, it would be realistic to assume that both countries produce rice. In this case, in a model of free trade, the amount of rice traded would depend on both domestic yield shocks as well as foreign shocks. For instance, it is expected that importing countries would decrease imports in response to positive domestic yield shocks and increase imports when there is a positive yield shock in the foreign country. As imports feed into consumption, we can also consider the consequences for this indicator of economic welfare. For both countries, consumption is expected to be positively related to both domestic and foreign yield shocks. In the extreme and unrealistic case of perfectly integrated markets, the source of the yield shock would not matter. A weaker hypothesis is that consumption depends positively on both domestic and foreign yield shocks. We now test these hypotheses.
Our data set on country production, area, and stocks is drawn from the US Department of Agriculture. To compute exogenous shocks, we smooth the yield series using the Holt-Winters double exponential method. The deviation of the smoothed series from the observation is defined as the yield shock. This is computed for every country. For every country, we also compute a foreign yield shock, which is the production weighted average of the yield shocks in each of the countries constituting the rest of the world.
To examine the potential of trade, the correlation between domestic yield and foreign yield shocks is worth considering. When there are adverse shocks to both domestic and foreign yields, trade cannot be of much help. To assess the probability of such outcomes, we slice domestic and foreign yield shocks into three categories: a high negative shock, when the shock is one standard deviation below the mean; a high positive shock, when the shock is one standard deviation above the mean; and a mid-range shock, when the yield deviation is within one standard deviation of the mean. This is done for every country and for every year in the sample. The cross-tabulation of these shocks for all countries in the sample is displayed in Table 2. Table 3 contains these cross-tabulations for the major countries that make up world production and trade: Bangladesh, the PRC, India, Indonesia, Iran, Malaysia, Nigeria, Pakistan, the Philippines, Saudi Arabia, Thailand, Viet Nam, and the US. The results show that in only about 3% of the cases for the entire sample and in about 2% of the cases for the major countries, low domestic yields are accompanied by low foreign yields as well. This means that except for these instances, trade, in principle, should work well in the overwhelming majority of circumstances when domestic production shortfalls are offset to some extent by higher output elsewhere, and vice versa. Yet the puzzle is that rice trade is considered unreliable relative to other grains.  Table 4 is a regression of the first difference in log of imports (as proportion of consumption) on the dummy variables for each of the categories in the cross-tabulations of Tables 2 and 3. The regression is based on the sample of all importing countries. As expected, the percentage change in imports is negative and the greatest in absolute value when the domestic shock is highly positive and the foreign shock is highly negative. This is the case when the demand for imports is at its minimum and the world supply is also at its lowest. Unsurprisingly, percentage change in imports is positive and maximal when the domestic shock is highly negative and when the foreign shock is highly positive. This is the opposite case when world supply is at its maximum and so is the demand for imports. These are instances when trade works in the expected direction. More surprisingly, imports as a proportion of consumption increase even when shocks are negative at home and abroad. In this case, world supply is low but import demand is high.
There is a clear pattern to the results. The percentage change in imports is less (or negative) when domestic shocks are highly positive; it is high and positive when domestic shocks are highly negative.
To see the cost of highly negative domestic shocks, consider a regression of the log change in rice consumption as a function of the dummy variables representing the combination of highly negative, mid-range, and highly positive domestic and foreign yield shocks. Table 5 shows the results for the entire sample of countries, not just importers. A second specification in the table adds lagged values of the dependent variable as regressors. 2 The impact of the shocks does not vary much between the specifications in terms of the sign and significance of the coefficients. 2 Conventional fixed effects estimators (such as the within estimator) are inconsistent when lagged values of the dependent variable are used as regressors. We used the Arellano-Bond estimator which transforms the data into first differences and takes care of the correlation between the error term (first difference of the original error term) and the lagged first differences of the dependent variable by using higher-order lags of the dependent variable as instrumental variables (Arellano and Bond 1991).  Reading from the first specification, in the scenario of highly negative domestic and foreign yield shocks, rice consumption declines by 9%. 3 In the scenario of highly negative domestic shocks but highly positive foreign yield shocks, rice consumption declines by 4.5%. The difference in outcomes between these scenarios is a measure of the value of access to world markets. However, consumption declines in all the scenarios involving negative domestic yield shocks. Positive foreign shocks can compensate, but not fully. Earlier, we mentioned that reliance on trade could fail in 2% of the instances when negative shocks affect both domestic and foreign markets. But now it is apparent that rice consumption is vulnerable in all the scenarios involving negative domestic shocks. Such instances occur 12% of the time. Perhaps this is why rice markets are regarded as "unreliable" in the literature.
The flip side of these results is that rice consumption increases by 10%-13% in all the scenarios involving positive domestic shocks. Most strikingly, the increase in consumption in the scenario of positive domestic and foreign yield shocks (13%) is almost the same as in the scenario of positive domestic and negative foreign yield shocks (12.5%). The failure of trade to redistribute supplies in the latter scenario seems to be the reason why trade is not able to stabilize consumption in countries hit by negative domestic shocks even though world supplies are ample. Table 6 is the consumption regression for some of the Asian countries important in the world rice economy: Bangladesh, the PRC, India, Indonesia, the Philippines, and Viet Nam. Pakistan and Thailand are excluded. 4 Once again, the implied rates of consumption change do not vary greatly between the two specifications. Exports as a proportion of consumption are greater than 50% in both these countries. The vulnerability of domestic consumption to yield shocks would not be a major concern here. The regression uses country fixed effects. 3. The specification with lagged dependent variables has been estimated with the Arellano-Bond method using second-to sixth-order lags of the dependent variable as instrumental variables. Source: Authors' estimates. Table 7 compares the average percentage change in rice consumption in each of the shock scenarios for the entire sample and for the Asian sample. The common finding is that rice consumption declines are substantial and comparable in the scenario of negative domestic and foreign shocks. However, Asian countries seem to do better to arrest consumption declines in the other scenarios involving negative domestic yields. The most striking difference involves the positive domestic yield scenarios: the consumption growth in the Asian countries is lower than in the world sample. This could be due to exports or the build-up of domestic stocks. The latter seems more likely because, as in the world sample, the difference in consumption growth between the scenarios of positive and negative foreign shocks (given positive domestic shock) is small. Domestic stocks in turn may have enabled these countries to stabilize consumption when domestic shocks are negative. Yet, even this policy has not been successful when negative domestic shocks are accompanied by negative foreign shocks.

VI. POLICY RESPONSE
It is clear that negative domestic shocks occur when stabilization fails to take place. Access to world markets helps but even when foreign yields are high, consumption declines. These are reduced form results and the outcome of both trade and domestic stabilization policies. To understand how exogenous shocks are modified by trade and domestic policies, we consider the following regression model for country j and year t: where C is rice consumption; DY and FY are domestic and foreign yield shocks; DS and FS are the domestic and rest-of-the-world stocks, both as proportions of domestic and rest-of-the-world consumption, respectively, at the beginning of year t; and is a country fixed effect. Earlier, we explained how shocks were constructed.
In our data, the policy variable is the level of stocks in each country. Clearly, trade restrictions will have a direct impact on stocks. For each country, we construct a domestic stock variable and a foreign stock, which is an aggregate of the stocks in the rest of the world. We allow the coefficients of domestic and foreign yield shocks to vary with domestic stocks and foreign stocks. In particular, The results are presented in Table 8. Both domestic shocks and domestic stocks have a positive impact on the change in consumption, and are statistically significant as well. Foreign yields and foreign stocks are not significant. The interaction term involving domestic shocks and domestic stocks is significantly negative. This shows that domestic policies moderate the impact of domestic shocks. A more revealing approach is to use the classification of shocks into negative, midrange, and positive. This allows policies to interact with shocks in a nonlinear manner. In this approach, the domestic shock variable is represented by dummies for negative, mid-range, and positive shocks. Let us call these dummies N d , M d , and P d . The foreign shock variable is represented similarly. Let us call those dummies N f , M f , and P f . Both sets of dummies are interacted with domestic and foreign stocks. The results can be seen in Table 9. The omitted base category in the table is the combination of mid-range domestic and mid-range foreign yield shock. From the table, we see that when both domestic and foreign shocks are negative, the expected value of the dependent variable is -0.208 + 0.436DS + 0.327FS. Thus, both domestic and foreign stocks help in stabilizing consumption in this state. However, the effect of foreign stocks and, by implication, trade, is not significantly different from 0. The median value of domestic stocks as a proportion of consumption is 0.05. This means that its contribution in reducing the hit on consumption is about 2.2 percentage points. The 75-percentile level of stocks is 0.2, and at this level, stocks would arrest the decline in consumption by 8.7 percentage points. The mean level of the stock ratio when both shocks are negative is 0.14. This reduces the negative impact on consumption by 6.1 percentage points. The stock-to-consumption ratio would have to be 47.7% to fully wipe out the adverse impact of domestic and foreign shocks.
The median level of the foreign stock ratio is 0.21 and that can help in countering the adverse impact by 6.9 percentage points. However, as noted earlier, this effect is not precisely estimated.

VII. CONCLUDING REMARKS
There is considerable literature about world price volatility and the transmission of world prices to domestic prices. In this paper, we have taken a different route to assess stability and to examine the role of trade and domestic stabilization policies. For each country, we constructed exogenous domestic and foreign (i.e., rest of the world) yield shocks, and looked at their impact on rice imports and on rice consumption. We also considered how this impact was modified by domestic and foreign stocks.
If supply shocks are uncorrelated across countries, the global supply is essentially stable. Provided that there are no demand shocks, the global price is also stable. Importing countries would be able to import, whenever they need to, at a stable price. 5 Even if shocks are correlated across countries, as long as the correlation coefficient is less than 1, the global aggregate supply is a lot more stable than individual country supplies.
Although trade cannot be expected to play a strong role when the major producing and consuming countries are simultaneously hit by negative yield shocks, such a scenario obtains in only 3% of cases. In all other cases of negative domestic shocks, they could be at least partially neutralized by positive foreign shocks. This implies that in a world of free trade, consumption levels in individual countries would be stabilized. However, our study finds that this is not the case. In cases of adverse domestic shocks, consumption fails to be stabilized even when foreign shocks are positive; however, imports do peak. Thus, while trade does help in coping with domestic risks, it is unable to achieve full risk sharing. The flip side is that when domestic yield shocks are positive, consumption surges even when the shock in the rest of the world is negative. Therefore, it is clear that irrespective of foreign shocks, the principal concern for poor countries is to stabilize consumption when hit by negative domestic yield shocks. The frequency of such shocks is about 12%.
Domestic policies have played a greater role in stabilizing the adverse impacts of negative shocks. This could be because of the presumed "unreliability" of rice trade. Storage is expensive, however, and countries often tend to carry too much stock either because of extreme precaution or because these policies are captured by producer interests. Furthermore, reliance on domestic policies will continue to keep rice markets thin and promote market insulation policies similar to those that led to the rice price spike in 2007-2008. The positive development in the world rice market has been the greater volume of trade since the mid-1990s due to the export liberalization in India and the entry of Viet Nam into world markets. Can there be another shift upwards? Surpluses in the commercial rice exporting countries such as Thailand, Pakistan, and the US are already high. Exports are as high as domestic consumption in Thailand and Pakistan, while in the US, the ratio is close to 60%. That is why the thickening of the rice market had to depend on new exporters such as India and Viet Nam.
Between 2006 and 2008, Viet Nam's exports were consistently around 21% of consumption. However, Indian exports have varied between 2.5% and 6% of domestic consumption. Not only has India's contribution to world exports varied, but the surpluses have also been small relative to domestic consumption. Negative domestic shocks together with domestic policies can shrink these surpluses quickly. Similarly, in the other large rice-producing economies such as Bangladesh, the PRC, and Indonesia, the surpluses or deficits are small relative to consumption, and it is not clear whether they can be reliable contributors to global supplies in the future. Besides, climate change poses unknown perils to some of the major rice growing regions in Bangladesh and India.
In this sense, the rise of Viet Nam is reassuring to the long-term future of the world rice market, although the surpluses are not as large as in Thailand. While surpluses may continue to rise in Viet Nam, especially with rising prosperity, we might need to see the emergence of surpluses in other countries for the rice market to thicken. Myanmar and Cambodia are possible candidates for rice export. It does seem that a more reliable rice-trading system would have to await greater productivity increases in some of the key rice-producing regions of the world.