Finance and Economics Discussion Series: Accessible versions of figures for 2015-094

Input Linkages and the Transmission of Shocks: Firm-Level Evidence from the 2011 Tohōku Earthquake

Accessible version of figures


Figure 1. Index of Japanese Industrial Production: Manufacturing Jul.2010 – Jan.2012.

Figure plots the index of Japanese industrial production at a monthly frequency. The y-axis measures the index after removing the trend component using an HP-filter. The x-axis measures time at a monthly frequency, beginning at July 2010 and ending at January 2012. The figure begins at a value of around 0.03 and is flat for some months, with a slight upward movement for about 3 months. Then in the month of March 2011, the figure drops by roughly 15 percentage points on the graph, and then slowly recovers to its pre-drop level by around August-September of 2011. The figure is roughly flat thereafter.

Return to text.


Figure 2. U.S. imports from Japan and Rest of World, Jul.2010 – Jan.2012.

Figure plots two lines: the U.S. imports from Japan, and also the U.S. imports from the rest of the world. The y-axis measures these imports after removing the trend component using an HP-Filter. The x-axis measures time at a monthly frequency, beginning at July 2010 and ending at January 2012. The line representing U.S. imports from Japan begins at around 0.03 and has a very slight upward trajectory. The line falls by around 20 percentage points on the figure in the month of April 2011, and then gradually increases in the months that follow. By the month of September 2011, the line is roughly at its point prior to the drop. The line representing U.S. imports from the rest of the world begins at around 0.01 and is roughly flat for the whole period. There is a slight upward movement from December 2010 until March 2011, but this gradual increase is only roughly 5 percentage points in total.

Return to text.


Figure 3. U.S. industrial production: manufacturing and durable goods.

Figure plots two lines: the U.S. index of manufacturing production and the U.S. index of durable goods manufacturing production during the period of July 2010 to January 2012. The y-axis measures the index after removing the trend component using an HP-filter. The x-axis measures time at a monthly frequency, beginning at July 2010 and ending at January 2012. The line for manufacturing production begins at about 0.012 and is roughly flat until the month of April 2011. At that point there is a sharp fall of roughly 1 percentage point. The line remains at this low level for about 5 months before gradually returning to the level before the drop. The last few points of the line (corresponding to December and January) have a jump upwards of about 1 percentage point. The line for durable goods manufacturing production starts at a value of 0.017, slightly higher than that for manufacturing production. It is also roughly flat until April 2011, where it drops by an amount of 1.5 percentage points. The movement in the durable goods production following this drop is similar to the movement in the manufacturing production. The line remains at a lower level for roughly 5 months before gradually returning to the level before the drop , which is achieved following a jump upwards in the months of December 2011 and January 2012.

Return to text.


Figure 4. Density of firm-level exposure to Japanese imported inputs: by firm type.

Figure plots a measure of the density (relative frequency) of firms across a measure of the exposure to Japanese imported inputs in the year 2010. The figure plots this measure separately for two firm groups: Japanese affiliates and non-Japanese firms. The y-axis measures the density across each firm group, and ranges from 0 to 0.5. The x-axis measures the log exposure to Japanese imported inputs, and ranges from -12 to 0. The line for non-japanese firms begins at a very low density close to zero, and then gradually rises as the log exposure measure becomes less negative. It reaches a peak of roughly 0.018 at a value of -6 for the log exposure measure, and then gradually declines back to zero as the log exposure measure approaches zero. The line for Japanese affiliates only begins at a measure of log exposure of about -0.8, with a density that is very low, essentially zero. This line increases as the log exposure measure gets less negative, and then increases rapidly beginning at around -2, reaching a peak of roughly 0.45 at a log exposure measure of -0.5. It then rapidly declines as the log exposure measure approaches zero.

Return to text.


Figure 5. Dynamic treatment effects: Japanese firms.

The figure has two panels. The Panel A subtitle is “relative intermediate input imports of Japanese firms. The figure in panel A plots the relative intermediate input imports , in dollars, of Japanese firms relative to a control group. The y-axis is measured in millions of U.S. dollars and ranges from -6 to 2. The x-axis measures time at a monthly frequency relative to the month of the earthquake (March 2011) which is denoted 0. Thus the x-axis ranges from -4 to 9. The line for the relative intermediate inputs is around zero up until month zero. It dips down and reaches a trough of about -3 at month=3 and then gradually returns to near zero by the end of the period (period 9). Also included in panel A are lines corresponding to confidence bands surrounding the main line. These indicate that the drop is significantly different from zero for months 1 through 4 following the earthquake. Panel B has a similar look to Panel A, but the subtitle for Panel B is “relative north American exports of Japanese firms”. The figure in panel B plots the relative North American exports, in dollars, of Japanese firms relative to a control group. The y-axis is measured in millions of dollars and ranges from -3.5 to 1. The x-axis measures time at a monthly frequency relative to the month of the earthquake (March 2011) which is denoted 0. Thus, the x-axis ranges from -4 to 9. The line for the relative output is close to zero until month 1 where there is a large dip. The trough occurs at month=2 at a value of roughly -2. The line then recovers slowly back to the initial level of roughly zero, reaching it at month 6. It remains close to zero for the remaining periods.

Return to text.


Figure 6. Relative Inputs and Output (Proxy) of Japanese Firms: Fraction of Pre-shock Level.

The figure plots the relative imported inputs and output of Japanese firms relative to a control group in the months preceding and following the March (2011) earthquake. The figure takes the lines from the previous figure (figure 5) and converts them into fractions by dividing through by the average level in the months before the shock. The y-axis is measured as a fraction of pre-shock levels and ranges from -0.6 to 0.1. The x-axis measures time at a monthly frequency relative to the month of the earthquake (March 2011) which is denoted 0. Thus the x-axis ranges from -4 to 9. The two series: relative imported inputs and relative output, track each other very closely and follow the path specified in figure 5. Both lines track close to zero until period 0, where both fall sharply, reaching values of roughly 45 percent below the pre-shock level at period 2 and 3. Then in the subsequent periods the lines return back to zero, reaching it at month of 6 or 7.

Return to text.


Figure 7. Assessing the Output Proxy Using Monthly Automotive Production.

The figure plots the line of relative output of Japanese firms from figure 6, in addition to another measure of output that is specific to the auto industry for Japanese firms. The y-axis is measured as a fraction of pre-shock levels and ranges from -0.6 to 0.1. The x-axis measures time at a monthly frequency relative to the month of the earthquake (March 2011) which is denoted 0. Thus, the x-axis ranges from -4 to 9. The two lines track each other closely. Both lines track close to zero until period 0, where both fall sharply, reaching values of roughly 50-55 percent below the pre-shock level at period 2. Then in the subsequent periods the lines return back to zero, reaching it at month of 6 or 7. The line corresponding to auto output falls slightly before and by slightly more than the line for the output proxy from figure 6.

Return to text.


Figure A.1 Geographic Distribution of Earthquake Intensity and Affiliate Location.

The figure shows a map of the northeast region of Japan that was most affected by the earthquake in March 2011. The map shows colors that correspond to the areas where the earthquake was the most severe. This is generally in the north and east part of the map and close to the Pacific ocean. On the map are also a series of dots that represent the locations of affiliates in Japan for which there is a firm-level match to United States manufacturing.

Return to text.


Figure A.2 Evidence of Potential Spillovers: Relative Non-Japanese Imported Inputs of Japanese Firms: Fraction of Pre-shock Level.

This figure reports the imported inputs from non-Japanese countries of Japanese firms in the U.S. relative to a control group in the months surrounding the March 2011 Japanese earthquake. The y-axis is measured as a fraction of pre-shock levels and ranges from -0.6 to 0.1. The x-axis measures time at a monthly frequency relative to the month of the earthquake (March 2011) which is denoted 0. Thus, the x-axis ranges from -4 to 9. The line hovers around 0 until month zero when it drops substantially, hitting a trough of -0.6 in month 3. There is some variation in the months that follow, but the broad trend is a recovery that almost returns to zero by the end of the figure (period 9).

Return to text.


Figure A.3 Japanese Products: Average Distance from Benchmark Cost Shares: JPN Multinationals.

This figure reports the average total (absolute) deviations from a benchmark measure of a firm’s cost shares across input products from Japan. The y-axis is measured as the average distance from the benchmark cost shares, at a 3-month moving average. The y-axis ranges from 0 to .25 The x-axis measures time at a monthly frequency and ranges from January 2009 to December 2011. The line begins at roughly 0.22 and then falls until the months of May and June 2010 – the benchmark period that was used as a reference point. The trough occurs at a value of roughly 0.08. After this date the line begins to increase again, with only minor bumps and troughs, until hitting roughly 0.15 in month December 2011. There are no sharp movements in this measure surrounding the 2011 earthquake.

Return to text.


Figure A.4 Automotive Production, Inventory, Sales by Firm Type, Distributed Lag Model.

This figure has four panels, in 2 row by 2 column format. The top left panel corresponds to North American production, the top right panel corresponds to U.S. sales, the bottom left panel corresponds to U.S. inventory, and the bottom right panel corresponds to Imported Sales and Inventory. With the exception of the bottom right panel, the other three panels have two lines per panel, which are the non-Japanese and Japanese automakers. The bottom right panel repots Japanese import inventories and Japanese import sales. The y-axis in all panels corresponds to the deviations from trend and range from -0.75 to 0.6. The x-axis corresponds to the months relative to the earthquake, and range from -3 to 8. In all panels, the line corresponding to non-Japanese automakers stay flat and close to the zero. At most the small movements up and down may reach a value of 0.08. In all cases the lines corresponding to the Japanese automakers fall in the month of 1 relative to the earthquake. The magnitude of the fall depends on the series considered. For North American production, the fall is almost to -0.5 and returns to zero at month = 6. For U.S. sales the line drops to -0.25 and again returns to zero in month 7. For U.S. inventory, the drop is less sharp and hits a trough at month = 3 at a value of -0.2, and just barely returns to zero by month 7 or 8. For the bottom right panel, the Japanese import sales fall in month 2 at a value of -0.2 and then return to zero by month 6. The Japanese import inventories fall more gradually, and hit a trough of -0.4 by month 3, and then gradually return to zero.

Return to text.


Figure A.5. Relative Inputs and Output (Proxy) of Japanese Firms (Reduced Sample) Logged, HP-Filtered.

There are two panels to this figure, both of which show the deviations from trend of imported inputs and output for Japanese firms relative to a control group. Panel A shows results without any weighting, where Panel B shows results that have been weighted according to size. For Panel A there are two lines: the output and imported inputs measure. The y-axis measures deviations from trend and ranges from -0.2 and 0.15. The x-axis measures time at a monthly frequency, relative to the month of the earthquake. The lines are somewhat noisy in the months preceding the earthquake, going up and down in-between -0.05 and 0.05. In month 1 both the imported inputs and output proxy fall, reaching a value of -0.2. The lines then begin to increase back to zero (the output proxy with some jumps up and down) and reach zero again in month 5 or 6. The output proxy has a slight bump up reaching 0.08 in month 6 before declining somewhat. Panel B has the same two lines. The y-axis measures deviations from trend and ranges from -0.6 to 0.2. The x-axis measures time at a monthly frequency, relative to the month of the earthquake. The two lines remain around zero in the months preceding the earthquake, and both drop in month 1. The output measure reaches a trough of -0.59 in month 2 and then returns to zero in month 5, continuing to climb to almost 0.15 in month 9. The imported inputs line reaches a trough of -0.32 in month 3 and then increases back to zero at month 6.

Return to text.


Figure A.6 Dynamic Treatment Effects: Relative Japanese Exports of Japanese Firms.

The figure plots the relative Japanese exports of Japanese firms in the U.S., relative to a control group in the months surrounding the 2011 earthquake. The y-axis measures millions of U.S. dollars and ranges from -0.8 to 0.6. The x-axis measures time relative to the earthquake and ranges from -4 to 9. The line wiggles around between -0.1 and 0.15 without any clear patterns. There are also lines corresponding to a confidence interval, and these lines encompass zero for all periods in the figure.

Return to text.


Figure A.7 Density Estimates of Elasticities Across Bootstrap Samples.

This figure has four panels: Panel A reports density estimates (relative frequencies) that correspond to the Materials elasticity estimates for Japanese multinationals and non-Japanese multinationals, Panel B reports density estimates that correspond to the capital/labor elasticity estimates for Japanese multinationals and non-Japanese multinationals, Panel C reports density estimates for the materials elasticity estimates for non-multinationals and all firms in the sample, and Panel D reports density estimates for the capital/labor elasticity for non-multinationals and all firms in the sample. The y-axes measure the density and range from 0 to 15. The x-axis measures the relevant elasticity estimates from the bootstrap samples and range from 0 to 1. Both Panel A and Panel B report two lines: one for Japanese multinationals and one for non-Japanese firms. In Panel A, the density of the materials elasticity for Japanese firms begins at roughly 4 at point 0 and then hovers around that value until declining at an elasticity value of roughly 0.25. It quickly drops to zero and remains at zero for the remainder of the elasticity values. For the non-Japanese firms, the density values are mostly around zero, with the exception of a spike that occurs between the elasticity values of 0.58 and 0.67 (the spike hits a value of roughly 12). For Panel B, the density of the capital/labor elasticity for Japanese firms begins at roughly 8 and then increases to about 10 at an elasticity range of 0.05. Then the density falls quickly and hovers around zero from the elasticity ranges of 0.1 until 1. For the non-Japanese firms the picture looks almost identical. Both Panel C and Panel D report two lines: one for non-multinational firms and one for all firms in the sample. In Panel C the density for materials elasticity measure for non-multinationals begins close to zero and begins to increase at an elasticity value of 0.3, reaching a peak of 5 at roughly 0.4, and then slowly declining to go back to zero at around an elasticity value of 0.6. The materials elasticity measure for all firms begins at zero and starts increasing around an elasticity value of 0.2, hitting a density value of 2 around 0.4, and then increasing again to a range of just over 5 at an elasticity estimate of 0.6. It quickly declines back to zero thereafter. In Panel D the non-multinational line begins at a density of roughly 4 around an elasticity estimate of zero. It slowly goes down to zero by the elasticity estimate of 0.2. For all firms in the sample, the density shoots up to over 15 directly after 0 and then quickly returns to zero for the remainder of the estimates values.

Return to text.