El profesor-tutor virtual en el trabajo colaborativo
4.3 Formación para el desarrollo de competencias en la formación virtual y en CSCL
The estimation of the production function parameters allows measures of productivity and markup ability to be constructed. The productivity analysis presented in this subsection follows Olley and Pakes (1996) and is similar to Pavcnik (2002).8 However, the productivity measure is based upon a plant’s quantified measure of output instead of deflated revenue. The analysis of markup ability is conducted in a manner similar to the productivity analysis. The plant-level productivity measures are created by subtracting the expected level of output from the quantified output measure of the plant. This creates the productivity measure for plantiat timet as:
ωit = ˇqit−βˆslsit−βˆulitu −βˆkkit. (3.20) Olley and Pakes (1996) note that the use of (3.20), instead of the measure
ωit =φit−βˆit,
which is defined in (3.17), is advantageous because this productivity measure can be created for all observations in the sample, instead of only those observations where investment is greater than zero.9 Additionally, this measure allows the creation of productivity measures for the year 1996, which was excluded from the production function parameter estimation due to lack of plant exit information.
The plant- and time-specific productivity measure allows aggregate levels of each to be constructed on an annual basis. The aggregate productivity measure,Wt, is constructed as a weighted average of each plant’s timetproductivity using the plant’s share of the industry’s
8
Pavcnik also subtracts a base plant’s productivity from the RHS of (3.20). As Pavcnik notes, similar methods of productivity analysis have also been utilized by Caves, Christiansen, and Tretheway (19881), Klette (1996), and Aw, Chen, and Roberts (2001).
output,sqit, as the weighting scheme: Wt= nt X i=1 sqitωit = ¯ωt+ nt X i=1 (sqit−¯stq)(ωit−ω¯t) (3.21)
where the bars represent the mean of all plants at time t. Similar to Pavcnik (2002) and Olley and Pakes (1996), aggregate productivity is decomposed in the second portion of (3.21) into unweighted average productivity and the covariance between industry share and plant productivity. The creation of this covariance term allows intraindustry changes in output relative to productivity to be examined. If the covariance term is increasing, then output is shifting towards more productive plants.
Tables D.4-D.7 present the results of the aggregate productivity measure as well as its decomposition. In this table, results are standardized with the 1979 value equal to zero. Aggregate productivity declines in all industries over the entire time period examined. How- ever, the direction of changes in aggregate productivity varied within the 1979 to 1996 time period. Most notably, aggregate productivity increased in all four industries over the 1979 to 1981 time period. Similarly, all industries experience a drop in productivity in 1982. This decline in productivity corresponds with a continued fall in GDP, which, in conjunction with rigidities of plant choice in labor and capital, likely led to the underutilization of resources. Likewise, as shown in Table D.8, all of the industries experienced an increase in their capital to output and labor to output ratios during the recession. This evidence further supports the notion that excess capacity stemming from input rigidities led to the drop in productivity during the recession. Following the recession, productivity maintained levels exceeding its 1979 value until the mid-1980s. However, despite the output growth in the industry in the mid-1980s and 1990s, the productivity measure in each industry experiences a decline.
Prior work has examined the evolution of productivity using the 1979-1986 subsample of the data. Pavcnik (2002) uses the estimation approach of Olley and Pakes (1996), henceforth OP, to analyze the evolution of productivity. Investment is used as a proxy for productivity in a manner similar to that in this paper. However, plant-level markups are not addressed. Levinsohn and Petrin (2003), henceforth LP, develop a methodology that instead utilizes a
plant’s materials usage as a proxy for productivity. An aggregate productivity measure is created by applying each of these methods to the data used in this paper. These measures are displayed in Tables D.4-D.7. A comparison of productivity across estimation methods shows that each yields substantially different results. The LP estimation methodology shows an increase in productivity over the time period that is greater than either of the other two methods for all of the three-digit industries examined. Estimates from the OP methodology are substantially lower than the LP estimates, but show mild increases in productivity over the time period. The productivity measure calculated using the methodology created in this paper is even lower than the OP method for all industries except ISIC 372. This difference stems from the ability to address intraplant changes in price across time. The results suggest that these markups affect the aggregate productivity measure in all of the industries except 372.
The aggregate productivity measure examined above does not address the issue of resource reallocation within an industry. More notably, the unweighted-mean productivity measure increases substantially in all of the industries. However, these productivity gains are over- shadowed by the reallocation of output share away from the more productive producers in the late 1980s. The decomposition of the aggregate productivity can be used to examine this reallocation. The previously described covariance measure is displayed in Tables D.4- D.7. In all four industries, the covariance measure begins a downward shift in the mid-1980s. This supports the notion that the recession had a disciplinarian effect on the market. Plants with higher productivity levels were rewarded with greater market share. However, this ef- fect was short-lived. The period of growth following the recession allowed plants with lower productivity levels to regain the market share lost in the early 1980s.
By directly addressing plant-level markups in the estimation method, the aggregate pro- ductivity measure created in this paper contrasts with those measures created by previous estimation methods. The estimation method in this paper eliminates the effects of plant- level price changes on productivity by creating a plant-level measure of physical output. The use of the output measure in the estimation of productivity eliminates the changes in plant- level prices that may otherwise influence an aggregate productivity measure. If output is
reallocated towards plants with higher markups, but such an reallocation is not addressed in the estimation process, then the perceived gains will influence the aggregate productivity measure.
In order to create the aggregate markup ability of plants in an industry, a measure similar to that used to examine productivity,Mt, is created by taking the weighted average of plant- level markup ability using the share of industry quantified output as the weighting scheme. The markup measure,λi,is held constant across time. Therefore, this measure cannot be used to examine changes in markups. However, each plant’s share of the industry’s output does change over time. Thus, the changes in the distribution of output among plants of varying markups can be examined.This measure is also decomposed into an unweighted average and covariance as Mt= nt X i=1 sqitλit= ¯λt+ nt X i=1 (sqit−s¯qt)(λit−¯λt). (3.22)
Similar to aggregate productivity, if the covariance is increasing, then larger levels of output are being produced by plants with higher markups. The aggregate markup measure is shown in Table D.9. The aggregate measure provides evidence that markups dropped in three of the industries during the time period leading into the recession. This is indicative of the pressure placed upon plants leading into the recession. However, the aggregate measure is generally increasing for all of the industries except ISIC 381.
The liberalization of trade barriers should also impact the ability of plants to mark up their products. Using data from five developing economies, including Chile during the same years examined in this paper, Roberts and Tybout (1996) find that price-cost margins are negatively correlated with trade exposure. Foreign competition reduces the ability of plants to mark up their final product over their costs.
The results in this paper support this notion brought forth by Roberts and Tybout (1996), but also show that exports play a role in determining the allocation of output. Table D.9 shows that the sign of the covariance measure depends on an industry’s trade orientation. The covariance remains negative in most years for import-competing industries, indicating that market share was reallocated towards the lower-markup producers. The export-oriented
industries have a positive covariance in most years, which indicates that plants with higher markups gained market share. If the exporters in an industry have higher markups relative to their non-exporting peers, and exports relative to output for domestic consumption increases, then these exporters would gain in market share. The positive covariance measure during the recession in conjunction with the increase in each of these industries export share (shown in Table D.2) provides evidence to support such a notion.
The reallocation of output across plants of differing markups also has an impact on mea- sures of industry output. The impact of the recession on revenue can be seen in Figure D.4. This figure compares each industry’s quantified measure of output with deflated revenue over the 1979-1996 period. The time t quantified measure of industry output, ˇQt, is constructed as: ˇ Qt= Nt X i=1 ˇ Qit= Nt X i=1 Mitβm,
where ˇQit represents the non-log quantified output of firm iat time t. The values are stan- dardized so that each represents a percentage change from 1979. These graphs show that the deflated revenue measure tends to understate the magnitude the recession. The growth of the quantified measure is greater in all industries except Food Processing. The difference in the quantified measure of output and deflated revenue stems from the varying markups in the industries. The aggregate markup measure falls during 1982 in all industries except ISIC 372, which experiences a large increase in 1983. As noted earlier, the estimation method assumes that λi is fixed over time. The changes in the aggregate markup stem from the reallocation of output within the industries. If output is reallocated to plants with higher markups during the recession, then the deflated-revenue measure of output would result in higher levels than the quantified measure. While a recession might lead to shift towards lower markup goods, an alternative explanation also exists. As domestic demand declines, exports compose a large portion of industry output. If high markup goods are exports, then the output-weighted measure would increase.
While the use of materials provides a proxy for the quantity of output under the assump- tion that raw materials are more homogeneous in nature than manufactured output, such a
technique is not immune from its own problems. The large increase in the quantified measure (parts b-d of Figure D.4) is driven by a drop in the raw materials price index during 1996. The use of this price index to deflate materials is likely responsible for the overstatement of the quantified measure of output in 1996.
The examination of the aggregate measures of productivity and markups provides evidence that supports several conclusions. The recession combined with input rigidities led to a fall in productivity. Despite this decline in productivity, the recession did lead to a reallocation of output share towards the more productive producers. The expansionary period following the recession reduced pressures on firms leading to an eventual decline in productivity. The markup measure shows a drop in markups of three of the industries during the recession, but these industries show a general trend of increasing markups throughout the sample period. More notable, however, is that the sign on the covariance indicating intraindustry reallocation of output corresponds with an industry’s trade orientation. A shift towards plants with higher markups occurs in the export-oriented industries, whereas output in the import competing industries moves to the lower markup producers. These results support previous findings that trade exposure leads to a reduction in markups.
This subsection has provided an analysis of productivity and markups during the 1979- 1996 period. The estimation method used in this paper results in aggregate productivity measures that are substantially different than previous methods. However, these differences may be partially attributed towards the reallocation of output to plants with a greater ability to markup their output. Although the trade-orientation of an industry provides some expla- nation of the evolution of markups within an industry, the impact of the recession plays a primary role in determining the evolution of the industries. The next subsection discusses the influences of international pressures on productivity in the industries examined.