The welfare cost of speculation during Kenyan trade reforms

While the theory of economic behaviour in the absence of credibility is fairly advanced, it has not been empirically tested. This paper provides such a test. More specifically, we quantify the social cost created by incredible trade reforms using Kenyan data.' The theory can be summarized by examining how private consumption and production decisions differ from those under perfectly credible policy reforms. These deviations may be due to either socially costly private action or a lack of desirable action. In consumption the welfare cost can result from non-optimal intertemporal substitution as private agents consume more today in anticipation of a higher future price (Calvo 1987). Similarly, consumers may accumulated speculative inventories of durable imports (Calvo 1988). In production, a lack of credibility is likely to encourage inaction, such as refraining from reallocation of factors of production to the export sector (Rodrik 1989), or deferral of investment until uncertainty concerning the future trade policy has been resolved (van Wijnbergen 1985). While waiting, entrepreneurs may accumulate excess foreign savings or speculative stock of imported inputs in anticipation of reversal of the reform. Our empirical analysis concentrates on speculative accumulation of stocks of durable imports and deferral of investment in Kenya during 1975-91. There are two necessary conditions for credibility: (i) time-consistency and (ii) macro compatibility of the reform, As suggested by Engel and Kletzer (1987) and Froot (1988), both these conditions are likely to be probabilistic rather than on-off situations. In the former case, private agents may not be certain about the true intentions of the liberalizing government and attach a subjective probability to default of the reform. In the latter case the possibility of adverse external shocks may undermine the permanence of liberalization. In the absence of sufficient political will to devalue, the only way the government can respond to a deterioration in the terms of trade is to reverse the reform. This possibility affects the expectations of private agents concerning the future price of importables and their intertemporal behaviour. If the public has rational expectations, it perceives the objective probability of collapse, whereas in the absence of rational expectations it will have to rely on a subjective probability. Regardless of the degree of uncertainty (including its absence), incredibility creates a social cost. Time-inconsistency may arise for several reasons. First, if the government is of a redistributive type, and factor mobility is costly, it is not able to credibly commit itself to the optimal policy of free trade. The public knows that the government will subsequently have an incentive to impose a tariff (explicit or implicit) and distribute the receipts to its favoured group (Staiger and Tabellini 1987). Second, time-inconsistency may result from conflicting goals of the public and the government, for example, regarding public spending (Daveri 1989). The private sector knows that the government will try to raise its revenue by unexpected inflation, once people have built up their real cash balances. As the increased revenue is used for its self-concerned spending, the government cannot commit itself to free trade. The presence of conflicting goals is equivalent to a distortion which interacts with another distortion, i.e. trade restrictions. If one distortion remains, it may not be optimal to remove the other. Third, there may be uncertainty whether the government is a true liberalizer or a redistributive one. In other words, uncertainty prevails over whether or not the time-consistency problem arises. As indicated above, this approach makes the lack of credibility probabilistic. Finally, the government can be endogenized as a political process where a majority, for example, determines whether the reform will be undertaken at all. In this case, time-inconsistency may result in a reform that never takes off due to uncertainty about winners and losers (Fernandez and Rodrik 1990). The other necessary condition for credibility is compatibility of trade liberalization with macroeconomic policies, such as fiscal, monetary and exchange rate policy. Similarly, to the time-consistency problem, private agents may observe incompatibility with certainty due to misguided government policies, or incompatibility may be probabilistic due to uncertainty created by the possibility that adverse terms-of-trade shocks will occur in the future. Many African economies use import controls to manage their balance of payments. By definition, the endogenous trade policy rule implies that any liberalization is temporary, unless the positive shock, which initially prompted the reform, turns out to be permanent rather than temporary. If private agents have rational expectations concerning the type of the shock, they are also able correctly to anticipate the fate of the reform. As a corollary, if there is no change in the terms of trade when a trade reform is undertaken, compatibility has to be brought about by a simultaneous devaluation, or by a reduction in the money supply. Temporary foreign aid, for example, allows the government to buy back domestic currency with foreign currency, provided aid will not increase public spending. Unlike the necessary conditions, the sufficient conditions for credibility are much more difficult to identify, or to predict. Even if the two necessary conditions are being met, there is a continuum of other reasons that may create a credibility problem, such as the number of years to the next election, or an expected shift in the balance of political power. Expressed as a function of the probability of collapse, the total social cost of incredibility is a non-monotonic curve with kinks. First, the welfare cost in consumption is an increasing function of the probability of collapse with a jump. While the cost of sub-optimal intertemporal substitution is monotonically increasing in the subjective probability, the speculative holding of stocks is likely to be nil at low probabilities. When the probability reaches a critical level, there is a jump in the social cost when stocks are purchased, given the levels of tariffs and interest rate. Hence, below the critical level there are no costs related to inventories but there is sub-optimal intertemporal substitution, and above the critical level the cost of holding stocks increases monotonically. Second, when the probability increases, the anticipated return to capital in the export sector relative to the formerly protected sector becomes less attractive. The incentive to relocate resources diminishes, increasing the welfare cost monotonically in the probability up to another critical value at which no reallocation of factors takes place and the welfare cost reaches its maximum level. Third, the social cost of deferral of investment depends on the rate of return and the level of planned investment in physical plant. At low probabilities, and given the level of tariffs, the social cost would be zero since private investors will act according to the policy-maker's intentions and invest in exportables. A third critical value of the subjective probability can be identified at which the cost jumps up since investment is deferred by accumulating stocks of imports which yield a zero social return. Empirically, one method of identifying trade liberalization episodes is to examine when actual policy changes, such as shifts in import licensing practices, tariffs, or export incentives, were initiated and when they were stopped or reversed. When introductions and reversals of policy changes are frequent, this approach alone is not sufficient to determine whether or not a given trade regime has became more liberal. Changes in trade policy instruments do not necessarily surface at the aggregate level as they may be in opposing directions, influenced by other policies, and variations in economic conditions. A quantitative index of trade policy is therefore required to establish the net effect of changes in policy instruments, for example, on relative prices. How can we diagnose the less than perfectly credible episodes from a narrative of changes in trade policy instruments? In principal, a candidate for such an episode should pass a three-faceted test. The first part of this test is to examine whether or not we can identify it as an episode by using a quantitative measure of trade policy.3 Second, we have to ascertain that at least one of the sufficient conditions for incredibility is being met, i.e. the reform is perceived to be either macro incompatible or time-inconsistent. As pointed out earlier, the two conditions are not likely to be either on or off but the perceived probability of reversal lies somewhere between zero and one. The third test is to make sure that sufficiently little else was simultaneously going on in the macroeconomic front so that we can attribute the event in question to trade policy. Finding reliable quantitative criteria of trade policy is not an easy task. Yet, an objective criterion is needed if we wish to study the investment response to liberalization, or the effects of trade policy on exports over time, or across countries. For such studies we do have to know what type of orientation the country's trade policy had during the period in question. However, when we want to assess the social cost of incredible trade reforms, the information requirement is likely to be less than when investigating other types of private responses. This is because we are actually assessing the relatively immediate effects of changes in policy instruments on private behaviour, regardless of whether they will ever surface at the aggregate level. In other words, we are looking at how the private sector reacted at the government's attempt to influence relative prices, i.e. trade policy, regardless of how successful it was in this attempt. Moreover, an annual quantitative index of trade policy could even be misleading in some cases as an episode may begin and end within a year, averaging out in the index, but still inducing speculative action and a social cost. For our empirical analysis of the credibility problem and speculation in Kenya during 1975-91, it is therefore justified to define liberalization episodes based on the narrative of changes in trade policy instruments rather than on a quantitative measure, such as the implicit tariff index. The following episodes will be studied in this paper: (i) endogenous import liberalization during the coffee boom in 1976-78, (ii) exogenous import liberalization in 1980, (iii) systematic licensing introduced in 1982, and (iv) relaxation of import controls and subsequent emphasis on export promotion, effective since 1989.4 The nature of the third episode is ambiguous ex ante. Although access to imports of intermediate and capital goods was improved, it was carried out at the expense of imports of consumer goods so that this reform may have to be classified as tightening of import controls rather than liberalization. What can we say about the speed of liberalization during these episodes? The speed of the first episode, which was essentially a temporary reform under the endogenous trade policy rule, was adjusted to the speed of relaxation of the foreign exchange constraint. Bevan et al. (1990) have, however, shown in a computable general equilibrium model that the extent of relaxation of import restrictions was not sufficient as there was a considerable shift of resources to domestic manufacturing at the expense of exportables. The 1980 liberalization can be labelled as a big bang as the removal of non-tariff restrictions was fast and extensive, although there were also gradual elements to it, such as a preannounced timetable for shifting import items to less restrictive categories. The speed of the 1982 reform is somewhat ambiguous as, on the one hand, various schedules were reformed one by one, i.e. gradually, but, on the other hand, once the revision of a schedule was completed, it was put into use at once. The last episode had a much more gradual approach to import liberalization than the 1980 reform. Cautiousness did not, however, make it credible to the public, although authorities seem to have been more effective in prevention of private speculation. Before attempting to quantify the welfare cost caused by speculative private action in Kenya in response to trade liberalization episodes, we will assess the credibility of the reforms identified above by examining data on external reserves. Thereafter, three hypotheses will be tested. First, we will examine how uncertainty concerning the permanence of a more liberal trade policy affects inventories of imports held by private agents. The null is that stocks of imported consumer durables and storable intermediate goods go up shortly before private agents anticipate that the reform will be reversed. This is because the expected gain from holding stocks goes up when it is anticipated that the reform can not be sustained and that the future price of importables will increase. In practice, stockpiling can take various forms, such as holding excess import and foreign exchange licenses, placing imports in bonded warehouses, or storing actual imports which have been declared through the Customs. According to our second hypothesis uncertainty about the future policy makes investors wait and accumulate inventories of importables - in the absence of (legal) access to foreign savings - rather than invest in physical plant (or low-yielding domestic financial assets). Policy-makers expect that the new liberal trade regime (or a promise of it) will induce investment in the export sector instead of importables. The shift may not, however, take place due to waiting and temporary preference for more liquid assets until uncertainty has been resolved. Finally, potential links between the two hypotheses will be investigated. Accumulation of stocks can be expected to create an increase in the demand for credit, whereas waiting in investment is likely to cause an increase in the supply of loanable funds.' The rest of the paper is organized as follows. Section 2 will assess the credibility of the four liberalization episodes, identified from the narrative of changes in trade policy. Section 3 will present a taxonomy of the range of choices available to the private sector to hedge against anticipated reversal of the reform, including relevant data to test the above three hypotheses. Section 4 will discuss how private agents can finance speculative inventory accumulation, while Section 5 contains a detailed analysis of private responses to each of the four liberalization episodes.. Finally, a summary and conclusions can be found in Section 6.