65 FR 55005, September 12, 2000                                               
   


                                                A-583-827 
                                                AR 10/97-3/99 
                                                Public Document


MEMORANDUM TO:         Troy H. Cribb
                       Acting Assistant Secretary 
                        for Import Administration

FROM:                  Richard W. Moreland
                       Deputy Assistant Secretary, Group 1
                        Office of AD/CVD Enforcement    
         
SUBJECT:               Issues and Decision Memorandum for the Antidumping Duty
                       Administrative Review on Static Random Access Memory 
                       Semiconductors from Taiwan - October 1, 1997, through 
                       March 31, 1999
    
Summary


We have analyzed the comments of the interested parties in the 1997-1999
administrative review of the antidumping duty order covering static random
access memory semiconductors from Taiwan.  As a result of our analysis of the
comments received from interested parties, we have made changes in the margin
calculations as discussed in the "Margin Calculations" section of this
memorandum.  We also recommend that you approve the positions we have developed
in the "Discussion of the Issues" section of this memorandum.  Below is the
complete list of the issues in this administrative review for which we received
comments from parties:


1.    Facts Available
2.    Date of Sale for Certain Transactions Related to a Joint-Venture Agreement
3.    Unreported Cost Data           
4.    Ordinary Course of Trade                   
5.    Winbond's Cash Deposit Rate
6.    Yields 
7.    Variances
8.    Foreign Exchange Losses Related to Cash Transactions  
9.    Research and Development Costs
10.  Products Produced But Not Sold During the Review Period
11.  Bonuses
12.  Clerical Errors in Winbond's Calculations
13.  Constructed Export Price Offset


Background


On May 8, 2000, the Department of Commerce (the Department) published the
preliminary results of the administrative review of the antidumping duty order
on static random access memory semiconductors (SRAMs) from Taiwan.  See Static
Random Access Memory Semiconductors from Taiwan; Preliminary Results and
Partial Rescission of Antidumping Duty Administrative Review, 65 FR 26577 (May
8, 2000) (Preliminary Results).  The products covered by this order are
synchronous, asynchronous, and specialty SRAMs from Taiwan, whether assembled
or unassembled.  The period of review (POR) is October 1, 1997, through March
31, 1999, for two of the reviewed companies (i.e., G-Link Technology Inc. (G-
Link) and Winbond Electronics Corporation (Winbond)) and October 1, 1998,
through March 31, 1999, for the remaining company (i.e., GSI Technology Inc.
(GSI Technology)).
    
We invited parties to comment on our preliminary results of review.  Based on
our analysis of the comments received and our findings at verification which
were not incorporated in the preliminary results, we have changed the results
from those presented in the preliminary results of review.


Margin Calculations


We calculated constructed export price (CEP) and normal value using the same
methodology stated in the preliminary results, except as follows:


  
           We increased G-Link's probing costs by the largest percentage
difference
           observed at verification between the costs reported for particular
models and
           those recorded in the accounting system for the same products.  See
Comment 1;


  
           We excluded certain exchange losses and bad debt expense from our
           calculation of the financial expense factor for G-Link.  Our
general practice
           is not to account for these types of expenses in the calculation of
financial
           expenses.  Therefore, we reclassified these items to the proper
expense
           categories (i.e., general and administrative (G&A) and indirect
selling
           expenses) for the final results;

  
           In accordance with our practice, we included the following items in
the
           calculation of cost of production (COP) and constructed value for G-
Link
           because they were excluded from G-Link's reported costs: an offset
for scrap
           revenue received on the sale of SRAMs; inventory write-offs;
exchange losses;
           write-downs related to raw materials and work-in-process (WIP)
inventory; and a
           standard cost adjustment variance related to raw materials and WIP;


           We included employee stock bonuses in the calculation of COP and
           constructed value for G-Link because we found during verification
that G-Link
           had mistakenly excluded its employee stock bonus expense from the
reported costs;


           We included in G-Link's COP and constructed value certain research
and
           development (R&D) expenses related to a discontinued imaging
business and
           amortization expenses associated with obtaining technology in
accordance with
           our general practice;
  

           We adjusted Winbond's fabrication costs to account for differences in
           fabrication yield loss.  Specifically, we increased these reported
costs for
           all products by the percentage difference between the fabrication
yield loss
           observed at verification and the reported yields.  See Comment 6;


           We included the standard cost variance associated with "WIP on
hand" in the
           calculation of Winbond's COP and constructed value.  See Comment 7;


           We revised the calculation of constructed value for Winbond to
exclude an
           imputed amount of stock bonuses for 1999 because we confirmed at
verification
           that Winbond did not incur any stock bonus expense for that year. 
See Comment
           11; and

  
           We deducted U.S. warehousing expenses from the calculation of CEP for
           Winbond.  SeeComment 12.
         
Discussion of the Issues


Comment 1: Facts Available
 
The petitioner argues that the Department should base the final margin for G-
Link on adverse facts available.  According to the petitioner, the Department
discovered at verification that G-Link did not use its normal cost accounting
system to calculate the reported costs, although its normal system provides
sufficient detail to allow it to do so.  The petitioner argues that the
company's decision to ignore its normal cost accounting system and develop
costs from scratch resulted in significant distortions in the data reported. 
Specifically, the petitioner maintains that, in most of the instances in which
the Department found differences between the reported and actual costs at
verification, the reported costs were lower, often by substantial amounts.


According to the petitioner, the Department faced a similar situation in
Certain Cut-to-Length Carbon Steel Plate From Mexico: Final Results of
Antidumping Duty Administrative Review, 64 FR 76, 77-78 (Jan. 4, 1999) (Mexican
Plate).  In that case, the Department resorted to total adverse facts available
because it found at verification that the respondent had failed to use its
normal accounting system in developing its COP and constructed value data.  The
petitioner asserts that the facts in this proceeding are analogous to those in
Mexican Plate because: 1) like the respondent in Mexican Plate, G-Link claimed
that its normal cost accounting system did not break out costs on a product-
specific basis; and 2) as in Mexican Plate, the Department found at
verification that G-Link's cost accounting system not only provided sufficient
information to determine product-specific costs, but that these costs differed
significantly from those reported.


The petitioner argues that, by failing to report its actual cost data, G-Link
deliberately withheld information that was adverse to itself.  Thus, the
petitioner argues that the Department should base G-Link's margin on adverse
facts available.  As facts available, the petitioner contends that the
Department should use the highest rate found in the less-than-fair-value (LTFV)
investigation (i.e.,113.85 percent).


G-Link contends that its cost calculation methodology was both conservative
and reliable.  G-Link disagrees that it would have been able to calculate
product-specific costs using its normal cost accounting system because this
system does not track variances on a model-specific basis. (1) Therefore, G-
Link states that it had to develop the data reported in its cost response.  G-
Link maintains that, although it did not start with the standard costs
reflected in its cost accounting system and adjust these costs for particular
variances, its approach (i.e., calculating the actual cost to produce each
model) was equally reasonable because it yielded the same result.  G-Link notes
that its methodology was fully explained in its questionnaire response, and was
not "discovered" at verification as the petitioner alleges.


According to G-Link, the concern that G-Link has lowered its costs by not
starting with the standards in its accounting system is not valid. 
Specifically, G-Link asserts that the main difference between the actual and
standard costs observed at verification results from an adjustment to the value
of finished goods inventory. (2)  G-Link contends that, because this adjustment
has nothing to do with producing SRAMs during the POR, the company
appropriately excluded it from the reported costs.  As support for this
contention, G-Link citesNotice of Final Determination of Sales at Less Than
Fair Value: Stainless Steel Wire Rod From Spain, 63 FR 40391, 40402 (July 29,
1998) and Antifriction Bearings (Other Than Tapered Roller Bearings) and Parts
Thereof From France, Germany, Italy, Japan, Singapore, and the United Kingdom;
Final Results of Antidumping Duty Administrative Reviews, 62 FR 2081, 2117
(Jan. 15, 1997). 


G-Link maintains that, when the inventory adjustment is eliminated, its costs
were in fact overstated.  Consequently, G-Link asserts that there is no basis
for the Department to apply adverse facts available to it for purposes of the
final results.


Department's Position:


We agree that G-Link did not use its normal cost accounting system to derive
the reported costs. However, we found at verification that the reported per-
unit costs, with the exception of probing costs, were generally the same as
those recorded in G-Link's normal accounting system.  Thus, we disagree that
these facts warrant the use of the highest rate found in the LTFV investigation
as an adverse inference.
     
Although G-Link made mention in its section D response that it departed from
its normal cost methodology, it did not provide any details regarding this
departure, nor did it contact the Department, as instructed by the
questionnaire.  Consequently, we were unaware of the actual method used until
verification.  At verification, company officials acknowledged that the costs
reported in the company's section D response were not those reflected in the
normal cost accounting system.  Specifically, G-Link officials explained that,
while the company normally computes model-specific costs using a standard cost
accounting system (i.e., model-specific standard costs adjusted to actual costs
through the application of company-wide variances), G-Link based its reported
model-specific costs on actual values obtained from its financial accounting
ledgers.  We also noted at verification that G-Link's standard cost system,
which is used to prepare its audited financial statements, accounts for all the
physical characteristics identified by the Department.  Thus, we do not find
persuasive G-Link's claim that its normal cost accounting system is inadequate
for reporting purposes simply because it relies on company-wide variances.  We
note that the Department's practice calls for respondents to start with the
most specific level of variances kept in their normal books and records.  If
this level of detail is deemed insufficient, we then may go beyond the normal
system.  (See, e.g., Notice of Final Determination of Sales at Less Than Fair
Value: Hot-Rolled Flat-Rolled Carbon-Quality Steel Products From Japan, 64 FR
24329, 24351 (May 6, 1999). (3))  In this instance, we found at verification
that G-Link did in fact maintain the necessary information in its cost
accounting system to reasonably convert some of its company-wide variances to
product-group variances. (See memorandum from Stan Bowen, et al, to Neal Halper
entitled "Verification of the Cost of Production and Constructed Value Data,"
dated May 19, 2000 (G-Link Cost Verification Report).)  Hence, we note that G-
Link's normal cost accounting system was sufficient for the Department's cost
reporting purposes.


Section 773(f)(1)(A) of the Act requires that costs be calculated based on the
records of the exporter or producer of the merchandise, if such records are
kept in accordance with the generally accepted accounting principles (GAAP) of
the exporting country and reasonably reflect the costs associated with the
production and sale of the merchandise.  Thus, we accept the results from a
respondent's normal cost accounting system if it reasonably allocates cost data
to the subject merchandise.  In this instance, G-Link maintained its normal
cost accounting system in accordance with Taiwanese GAAP and it generated costs
that reasonably reflected the costs associated with the production and sale of
the subject merchandise.  Thus, we find that G-Link should have reported the
costs from its normal cost accounting system.  In instances with similar
findings, we have not relied on the respondent's per-unit costs calculated for
purposes of an antidumping proceeding.  (See, e.g., Mexican Plate at 64 FR 77-
78 (where the Department rejected the cost calculation method used by the
respondent to determine product-specific costs for its COP and constructed
value data files because it generated per-unit costs that differed
significantly from those maintained in its normal accounting records), and
Final Results of Antidumping Duty Administrative Review: Extruded Rubber Thread
from Malaysia, 63 FR 12752, 12760 (Mar. 16, 1998) (where the Department
rejected the cost calculation method used by the respondent to determine
product-specific costs for its COP and constructed value data files, and relied
instead on the costs and variances from the respondent's standard cost
system).) 


Our section D questionnaire clearly instructs the respondent to notify us in
writing before preparing its section D response if it does not intend to rely
on its normal cost accounting system. (See Questionnaire, Section D-III,
entitled "Response Methodology.")  We allow companies to deviate from their
normal cost accounting system when that system does not appropriately allocate
costs to specific products and only after consulting with representatives from
the Department.  (See Mexican Plate.)  As noted earlier, G-Link failed to
disclose adequately this departure from normal practice in its questionnaire
responses.  It was not until verification that company officials explained that
G-Link did not rely on its costing methods used in the normal course of
business.  As a result, we could not properly evaluate the cost calculation
method until verification.


During verification, we found that G-Link normally prepares its GAAP-based
financial statements using model-specific costs calculated from its cost
accounting system.  Since G-Link is mainly a design company, it subcontracts
third parties to perform all production processes except in-house probing.  To
develop the standard costs of each model, G-link relies on estimated
subcontracted processing costs specific to the model and calculates a standard
probing rate that is unique for each model.  At verification, we compared the
reported costs for several models to the product-specific standard costs
recorded in the normal cost accounting system. From this comparison, we found
that the reported actual model-specific subcontracted processing costs were
comparable to the standard model-specific subcontracted costs plus G-Link's
various manufacturing variances as recorded in its normal accounting system. 
The model-specific probing costs as calculated by G-Link's normal cost
accounting system, however, were found to be significantly different than the
reported costs.  (See G-Link Cost Verification Report at page 18 and attached
Cost Verification Exhibit 22.)  Specifically, we found that G-link has
historically allocated all of its probing costs based on each model's unique
physical characteristics.  In contrast, the response cost calculation method
simply assigned a single average probing cost to all products (both subject and
non-subject), despite known manufacturing differences among them.  Because of
this misallocation, G-Link's reported per-unit probing costs differ
significantly in some instances from the corresponding amounts reported in the
cost accounting system.


Section 776(a) of the Act provides that, if an interested party withholds
information requested by the Department, fails to provide such information in a
timely manner or in the form or manner requested, significantly impedes a
proceeding under the antidumping statute, or provides information which cannot
be verified, the Department shall use, subject to sections 782(d) and (e) of
the Act, facts otherwise available in reaching the applicable determination. 
In this review, we found that G-Link failed to provide necessary information in
the form and manner requested, and we also found that the submitted information
was inaccurate in certain instances.  Therefore, we conclude that pursuant to
section 776(a) of the Act, use of facts otherwise available is appropriate. 
However, we have not found it necessary to resort to total facts available
because much of G-Link's reported information was verifiable. (4)  In this
instance, we increased the reported costs for all products by the largest
percentage difference we found between the reported product-specific probing
costs and the product-specific probing costs according to the normal cost
accounting system.  We find that this adjustment to probing costs corrects for
the misallocation that results from the company's failure to rely on its normal
cost accounting system.


We have continued to rely on the reported subcontracted costs because record
evidence supports G-Link's assertion that the manufacturing variances make up
the reconciling difference between the reported and standard subcontracted
amounts.  In other words, we concur with respondent that the model's standard
subcontracted cost plus the various manufacturing variances will equal the
actual subcontracted amount reported for the POR. (5)  (See G-Link Cost
Verification Report and G-Link's case brief at attachments 1 and 2.) 
Consequently, we find that our adjustment to probing costs reasonably converts
the reported costs to the costs recorded in G-Link's accounting system. 


Nonetheless, although we have accepted G-Link's costs in this segment of the
proceeding, we have done so primarily because we verified that these costs were
the same as those recorded in the company's cost accounting system (except for
the probing differences identified above).  Our acceptance of G-Link's costs
here does not relieve G-Link of the requirement to rely on its normal
accounting system for reporting product-specific costs in future segments of
the proceeding.  Failure to do so may result in our re-evaluation of this issue.
 
Comment 2: Date of Sale for Certain Transactions Related to a Joint Venture
Agreement


During the POR, G-Link shipped certain SRAM models to the petitioner, Micron
Technology, Inc. (Micron), under a joint venture agreement.  G-Link argues that
these transactions should be excluded from the Department's analysis for
purposes of the final results because the date of sale for them was outside the
POR.  Specifically, G-Link argues that the Department should consider the
agreement date (i.e., December 21, 1996) rather than the invoice date to be the
date of sale because the agreement fixed all of the material terms of sale,
including both price and quantity.


G-Link asserts that the Department's practice is to treat the date of a long-
term contract as the date of sale where the material elements of the
transaction are set forth in the contract.  As support for this assertion, G-
Link cites Certain Corrosion-Resistant Carbon Steel Flat Products and Certain
Cut-to-Length Carbon Steel Plate From Canada: Preliminary Results of
Antidumping Duty Administrative Reviews, Intent To Revoke in Part, Intent Not
to Revoke in Part, and Rescission of Review in Part, 64 FR 45228, 45231 (Aug.
19, 1999); Notice of Final Determination of Sales at Less Than Fair Value:
Emulsion Styrene-Butadiene Rubber From Mexico, 64 FR 14872, 14879 (Mar. 29,
1999); and Certain Cold-rolled Carbon Steel Flat Products from the Netherlands:
Final Results of Antidumping Duty Administrative Review, 64 FR 11825, 11829
(Mar. 10, 1999).  According to G-Link, in this case both the transfer of the
intellectual property rights and the price for the goods to be shipped were
established prior to the POR, and only purchase orders against the contract
were issued thereafter.  Thus, G-Link argues that these sales should be
disregarded for purposes of the final results.


According to the petitioner, G-Link's claim that the terms of sale for these
sales were set in a long-term contract is not supported by the record. 
Specifically, the petitioner asserts that the terms of sale were not set by the
document proffered by G-Link (i.e., the Joint Development Agreement), but
rather in the another document which is not on the record of this proceeding.  
In addition, the petitioner notes that the Joint Development Agreement does not
fix the price, and thus the date of this agreement cannot be used as the date
of sale for subsequent purchases under it.


Indeed, the petitioner argues that there is no basis for departing from the
Department's normal practice of using the date of invoice for all of G-Link's
U.S. sales.  The petitioner notes that G-Link reported the invoice date as the
appropriate date of sale.  According to the petitioner, the Department should
reject G-Link's belated attempt to revise its reported sales information.


Department's Position:


In accordance with 19 CFR 351.401(i), the Department's practice is to base the
date of sale on the invoice date unless we find that a different date better
reflects the date on which the respondent establishes the material terms of
sale.  In this case, the evidence on the record does not establish that the
contract date better represents the date of sale for G-Link's sales to Micron
because the contract in question does not set the price of these sales. 
Although the contract provides a formula for the calculation of the price, it
also contains a provision which allows the parties to alter the formula after
the contract date.  Specifically, the contract provides that the parties can
agree to any price that will enable Micron to resell the products in question
for a reasonable profit (as long as that price is above G-Link's cost).  


The Joint Development Agreement references a Purchase Agreement which
purportedly sets forth the time period and terms and conditions of the sales as
mutually agreed upon by the parties.  (See Exhibit H of G-Link's December 22,
1999, supplemental questionnaire response.) However, G-Link did not provide the
Purchase Agreement on the record of this proceeding and admitted that one does
not exist.  Therefore, the price of the merchandise sold by G-Link to Micron is
not known on the date of the contract because the contract makes clear that the
price can be set pursuant to future discussions between the parties.


Because G-Link's assertion that the price of the sales to Micron was set in
the Joint Development Agreement is not supported by the evidence on the record,
there is no basis for disregarding the presumptive date of sale identified in
our regulations.  Accordingly, we have continued to use G-Link's reported dates
of sale (i.e., invoice date) for sales to Micron, consistent with our approach
in Notice of Final Determination of Sales at Less Than Fair Value: Certain Cold-
Rolled Flat-Rolled Carbon-Quality Steel Products From Brazil, 65 FR 5554, 5574
(Feb. 4, 2000).


Comment 3: Unreported Cost Data


G-Link failed to report cost data for two products sold in the United States. 
According to G-Link, the Department erred in basing the margin for these
products on facts available because it did, in fact, report cost data for them.
Specifically, G-Link asserts that it provided cost data for products which were
physically identical to the products in question in all respects except for
speed.  G-Link maintains that, because there is no cost difference associated
with a difference in speed, the Department had all the costs necessary to
perform its margin calculations without resorting to facts available.


According to the petitioner, the Department appropriately used adverse facts
available to determine the margins for the models in question.  The petitioner
asserts that G-Link not only failed to establish that the cost for these models
was the same, but also that this assertion of fact would be subject to
verification if timely reported to the Department.  The petitioner argues that,
because this assertion was made after verification, it should be rejected on
the grounds that it is both untimely and unsubstantiated.


Department's Position:


In this review, we requested that all respondents provide cost data in the
same quarter as the quarter of their home market and U.S. sales or, when
production did not occur in that quarter, to provide cost data for the most
recent prior quarter in which production did occur.  Contrary to G-Link's
assertion, G-Link did not report cost information for the products in question
in the same quarter as the U.S. sale or in a prior quarter.  Moreover, G-Link
did not report cost information for products which were identical in all
respects except for speed in the same or a prior quarter.


Because G-Link has not explained why it was unable to provide the requested
data, we find that G-Link has failed to cooperate to the best of its ability in
complying with our requests for this information.  Accordingly, as adverse
facts available, we have used the highest non-aberrant margin calculated for
any of G-Link's other U.S. sales, consistent with our treatment of these
unreported costs in the preliminary results.  In selecting a facts available
margin, we sought a margin that is sufficiently adverse so as to effectuate the
statutory purposes of the adverse facts available rule, which is to induce
respondents to provide the Department with complete and accurate information in
a timely manner.  We also sought a margin that is indicative of G-Link's
customary selling practices and is rationally related to the transactions to
which the adverse facts available are being applied.  To that end, we selected
the highest margin on an individual sale in a commercial quantity that fell
within the mainstream of G-Link's transactions (i.e., transactions that reflect
sales of products that are representative of the broader range of models sold
by G-Link).


Comment 4: Ordinary Course of Trade for GSI Technology
 
GSI Technology contends that three sales in its third country database were
not made in the ordinary course of trade because they were not made in
commercial quantities.  Specifically, GSI Technology asserts that each of these
sales represented less than 0.06 percent of the total volume of reported third
country sales.  GSI Technology asserts that these sales had a disproportionate
effect on the preliminary dumping margin, because approximately half of this
margin was attributable to them.


GSI Technology argues that the facts in this case are identical to those in
the final determination in the antidumping duty investigation on canned
pineapple fruit from Thailand.  See Final Determination of Sales at Less Than
Fair Value: Canned Pineapple Fruit from Thailand, 60 FR 29553, 29562 (June 5,
1995) (Thai Pineapple).  According to GSI Technology, in that case the
Department found that: 1) 90 percent of a respondent's dumping margin was
attributable to a single sale to a third country; and 2) that sale represented
only 0.01 percent of the sales to that country.  GSI Technology asserts that,
based on these facts, the Department found that this sale was outside the
ordinary course of trade and excluded it from its calculations.  


According to GSI Technology, the analysis in Thai Pineapple was based on eight
factors: 1) customers; 2) terms of sale; 3) volume of sales; 4) frequency of
sales; 5) sales quantity; 6) sales price; 7) profitability; and 8) market
demand.  GSI Technology maintains that, although the Department found that a
majority of these factors must be met in order to conclude that a sale is
outside the ordinary course of trade, it was not necessary to consider every
factor.  According to GSI Technology, when its three sales are evaluated using
the above criteria, it is clear that they were not made in the ordinary course
of trade.  Specifically, GSI Technology notes that the average quantity of each
of these sales was less than two percent of the average quantity of sales of
the same control number in the third country (or, lacking sales of the same
control number, sales in the market); the total quantity was equal to or less
than 0.06 percent of the total volume of third country sales; and the gross
sales price and profit on these sales was significantly higher than the price
and profit of other sales of the same control number or in the same market.


GSI Technology asserts that the purpose of the ordinary course of trade
provision, as acknowledged in Thai Pineapple, is to prevent dumping margins
from being based on sales which are not representative and do not lead to fair
price comparisons.  Therefore, GSI Technology argues that the Department should
exclude these sales from its margin calculations for purposes of the final
results.


The petitioner argues that the sales in question were made in the ordinary
course of trade because they do not meet the majority of the criteria outlined
in Thai Pineapple.  Specifically, the petitioner notes that GSI Technology
discusses only three of the eight criteria in that case (i.e., sales price,
sales quantity, and profitability).  With respect to these criteria, the
petitioner maintains that GSI Technology frequently made sales in similar
quantities as the sales in question and the company made a number of sales of
two of the three models.   Indeed, the petitioner asserts that nothing in the
record indicates that these sales are "unique" in any fashion. Moreover, the
petitioner notes that GSI Technology's arguments regarding profitability are
misplaced because the company compared the profit margin on the three sales
with the average profitability of all other sales (including those that failed
the cost test and thus are outside the ordinary course of trade).  According to
the petitioner, this test is inappropriate because GSI Technology should have
excluded below-cost sales.  The petitioner claims that, when below-cost sales
are excluded from this comparison, two of the three sales at issue had profit
margins which were not dissimilar to the average profitability of the remaining
sales.


Finally, the petitioner asserts that there is no basis for any claim that the
three sales differed with respect to customers or terms of sale.  The
petitioner notes that GSI Technology sold all of its products to two third
country customers, and that the terms of sale for these transactions did not
differ from those for G-Link's other U.S. sales.


Thus, the petitioner contends that GSI Technology has not demonstrated that
the sales at issue were made in quantities or prices that are so dissimilar to
its other sales in that market so as to render them outside the ordinary course
of trade.  Consequently, the petitioner asserts that the Department should
reject GSI Technology's request that the Department exclude them from its
analysis for purposes of the final results.


Department's Position:


Under section 771(15) of the Act, the term "ordinary course of trade" is
defined as "the conditions and practices which, for a reasonable time prior to
the exportation of the subject merchandise, have been normal in the trade under
consideration with respect to merchandise of the same class or kind."  In
general, the Department's practice is to exclude sales made outside the
ordinary course of trade from its margin calculations, in order to prevent
dumping margins from being based on sales which are not representative of home
market or third country sales. See, e.g., Thai Pineapple, 60 FR at 29563.


The Department considers a number of factors when determining whether
particular sales are outside the ordinary course of trade.  For example, the
Department may exclude sales from its analysis which are of merchandise
produced according to unusual product specifications, merchandise sold at
aberrational prices, or merchandise sold pursuant to unusual terms of sale. See
the Statement of Administrative Action (SAA), H.R. Doc 316, Vol. 1, at 834
(1994).  Seealso, Thai Pineapple, 60 FR at 29563; and Final Determination of
Sales at Less Than Fair Value: Sulfur Dyes, Including Sulfur Vat Dyes, from the
United Kingdom, 58 FR 3253, 3256-57 (Jan. 8, 1993) (where the Department
excluded one home market sale which was: 1)  made at a greater quantity and
lower price than other sales to the same customer; and 2) out of line with the
prices and quantities of the vast majority of the respondent's other sales
transactions in the home market during the period of investigation (POI); 3)
made pursuant to an agreement which was concluded in a manner noticeably
different from the respondent's other sales during the POI).


In this case, we find that the sales at issue are not outside the ordinary
course of trade.  These sales were made: 1) in quantities and at prices which
were usual for the market under consideration; (6) 2) to one of GSI
Technology's normal third country customers; and 3) at the same terms of sale
as the majority of the company's other third country transactions.   In
addition, we find that GSI Technology made a number of sales of the identical,
or substantially similar, products as those at issue.


We note that many of GSI Technology's arguments address the quantity, price,
and profitability of a particular sale in relation to the quantity, price, and
profitability of sales of the same control number.  However, we find that this
analysis is not appropriate because there is nothing so unusual about these
products to set them apart from other products in the same database.  For
example, these products were not a special type of SRAM which could not be
reasonably compared to off-the-shelf models, nor were they SRAMs which had
markedly different performance characteristics.  Thus, we find that, while the
profit made on a particular sale may be higher than the profits made on other
sales of that model (especially when those other sales are below-cost), this
finding is not determinative of whether that sale is outside the ordinary
course of trade.  Indeed, when the profitability of the sales at issue is
compared to that of the other sales in the database, we find that the profit
margins were not unusual or unreasonable for the market under consideration. 
(See memorandum to the file from Shawn Thompson entitled "Ordinary Course of
Trade Data for GSI Technology in the 1997-1999 Antidumping Duty Administrative
Review on Static Random Access Memory Semiconductors from Taiwan," dated
September 5, 2000.)  Consequently, we have continued to include the three sales
at issue in our analysis for purposes of the final results.


Comment 5: Winbond's Cash Deposit Rate
 
According to the petitioner, the volume of Winbond's U.S. sales made during
this POR is significantly lower than the volume of U.S. sales it made during
the POI.  Specifically, the petitioner notes that Winbond's U.S. sales during
the POR amounted to approximately $1.3 million, whereas the volume of the
company's sales during the POI was almost $10 million. According to the
petitioner, this disparity in sales value between this review and the less-than-
fair-value (LTFV) investigation indicates that Winbond has manipulated its
sales to the United States in order to achieve a substantially lower duty
deposit rate. (7)  Consequently, the petitioner argues that the Department
should review whether Winbond's U.S. sales are bona fidetransactions,
particularly as measured against contemporaneous sales for comparable products
made by other SRAM producers, both in the U.S. and third country markets.  The
petitioner asserts that the Department should assign Winbond the cash deposit
rate calculated in the LTFV investigation if it concludes that the company's
U.S. sales are not an accurate reflection of its normal sales activity.  As
precedent for this action, the petitioner cites Certain Cut-to-Length Carbon
Steel Plate From Romania: Notice of Rescission of Antidumping Duty
Administrative Review, 63 FR 47232 (Sept. 4, 1998) (Carbon Steel Plate from
Romania), where the Department rescinded a review for the sole respondent after
finding that the company had no bona fide sales during the POR.  


The petitioner notes that, in the context of revocation determinations, the
Department has held that a comparison of a respondent's aggregate U.S. sales
quantity over the period of review to the quantity of its sales made in other
markets provides a reliable indicator of whether the respondent's U.S. sales
have been made in normal commercial quantities.  According to the petitioner, a
similar principle should apply here, in that the Department should require that
a respondent demonstrate that it made sales in normal commercial quantities
before making a radical adjustment to the company's cash deposit rate.


Winbond asserts that its sales are bona fide arms-length transactions made to
major institutional customers in normal commercial quantities.  Winbond notes
that the Department verified its U.S. sales as normal commercial transactions,
and asserts that there is no evidence on the record of this proceeding which
contradicts this finding.  Winbond points out that its customers are
multinational computer companies who would not deal with any supplier,
including Winbond, on anything other than a bona fide basis.


Winbond argues that there is no statutory, regulatory, or case precedent to
support the petitioner's position, because neither the statute nor the
regulations authorize the Department not to impose a new cash deposit rate at
the conclusion of an administrative review.  Indeed, Winbond notes that 19 USC
1675(a)(2)(C) directs the Department to use a determination in an
administrative review as the basis for liquidation of entries made during the
POR and to set the future cash deposit rates.  Finally, Winbond maintains that
the petitioner has not raised this issue in a timely manner because it failed
to do so prior to the preliminary results.


Department's Position:


Section 751(a)(1)(B) of the Act directs the Department to conduct a
administrative review upon request by an interested party in order to determine
the amount of any antidumping duty owed during the POR.  Sections 751(2)(A) and
(C) of the Act further direct the Department to determine the dumping margin
for each entry of subject merchandise during this period using the normal value
and export price (or CEP) for the particular company under review and to base
the cash deposit rate on this data.  Because Winbond requested this
administrative review in a timely manner and provided all data necessary to
calculate normal value and CEP, we have used this data to determine the revised
cash deposit rate applicable to Winbond.
   
Regarding the petitioner's argument that Winbond's U.S. sales were not made in
commercial quantities, and thus were not bona fide transactions, we disagree. 
As noted above, the Act does not require the Department to establish whether
sales in the U.S. market were made in commercial quantities prior to conducting
an administrative review and it is not the Department's practice to do so. 
Rather, the Act directs the Department to determine the antidumping duty due on
each entry of subject merchandise and to set the cash deposit rate accordingly.
Indeed, the Department may conduct an administrative review based on a single
U.S. sale and entry if we conclude that the sale in question was a bona fide
transaction.  See, e.g., Fresh and Chilled Atlantic Salmon From Norway; Final
Results of New Shipper Antidumping Duty Administrative Review, 62 FR 1430, 1432
(Jan. 10,1997) (Salmon from Norway) and Carbon Steel Plate from Romania, 63 FR
at 47234.


In determining whether a sale is bona fide, the Department considers whether
the transaction has been so artificially structured as to be commercially
unreasonable.  For example, in Carbon Steel Plate from Romania, we found that
the total costs borne by the importer in relation to the price that the
importer ultimately received for the product rendered the transaction
commercially unreasonable.  See Carbon Steel Plate from Romania, 63 FR at
47234.  Furthermore, in performing its analysis in this area, the Department
does not consider whether the respondent sold in commercial quantities in a
particular market, but rather whether the shipment was made on a normal
commercial basis.  For example, in Salmon from Norway, the fact that the
quantity involved in the transaction at issue represented a small fraction of
the respondent's total sales did not lead us to conclude that the sale at issue
was not a bona fide transaction.  See Salmon from Norway, 63 FR at 1432.


In this case, Winbond submitted a sales database containing several hundred
transactions.  We verified the accuracy of this data at Winbond's U.S. sales
offices.  During this verification, we found no evidence to suggest that the
transactions in question were not legitimate U.S. sales made on a normal
commercial basis.  Specifically, we found that these transactions were made in
the ordinary course of business to a number of large corporate customers who
were not affiliated with Winbond.  Given these circumstances, we find that
Winbond's U.S. sales were bona fidetransactions and, as such, we have used them
to determine Winbond's cash deposit rate for purposes of the final results.


Comment 6: Yields 
  
The petitioner argues that Winbond's methodology for calculating yields at the
fabrication stage distorts the company's manufacturing costs.  Specifically,
the petitioner asserts that Winbond included certain data in both the numerator
and denominator of its calculation, which resulted in an understatement of the
fabricated per-unit wafer costs for each quarter of the POR. (8)  According to
the petitioner, although Winbond claims that its methodology is an "industry
standard," the company was unable to demonstrate this at verification. 
Moreover, the petitioner maintains that this methodology has not been used in
any prior proceeding involving semiconductors.


The petitioner asserts that there is no information on the record that would
allow the Department to correct Winbond's fabrication yields.  Furthermore, the
petitioner argues that, even if the Department were able to do so, this would
be inappropriate because it would reward Winbond for failing to disclose its
methodology prior to verification.  Rather, the petitioner argues that the
Department should adjust the reported yields using adverse facts available.  As
adverse facts available, the petitioners assert that the Department should
apply the highest percentage difference found at verification to all of
Winbond's fabrication costs.


Winbond maintains that it accurately reported its fabrication yields.  Indeed,
Winbond asserts that the petitioner's suggested methodology (i.e., excluding
the data in question from the numerator and denominator of the calculation)
would distort the company's yield data to the extent that it had any impact on
the calculation at all.  Moreover, Winbond maintains that the petitioner's
statement that the Department has not used this methodology in prior
proceedings is incorrect because Winbond used (and the Department accepted) an
identical methodology for calculating yields in the LTFV investigation. 
Further, Winbond asserts that its practice is fully in accordance with process-
based cost accounting principles and, thus, no adjustment to the reported
yields is warranted.


Finally, Winbond notes that it has fully cooperated throughout all phases of
this proceeding. Consequently, Winbond asserts that, should the Department find
that an adjustment is necessary, limiting the adjustment to only the highest
difference found at verification would be unfairly adverse.  Rather, Winbond
asserts that the Department should adjust the reported costs for the
differences in fabrication yield found at verification for the specific
products examined and apply the weighted average of these differences to all
other products.


Department's Position:
            
We agree with the petitioner that Winbond's methodology for calculating yields
at the fabrication stage distorts the company's reported manufacturing costs. 
In Winbond's cost accounting system, the company applies predetermined
fabrication yield rates to its products.  To account for the difference between
the total actual and standard fabrication costs, Winbond applies a plant-wide
variance.  For submission purposes, Winbond used an adjustment factor to
convert its cost accounting system's fabrication yields to control-number-
specific yields.  While we agree with Winbond that it was appropriate to assign
the plant-wide yield rates on a more detailed basis than normally captured by
its accounting system, we disagree that the method used was proper.  To make
this adjustment, Winbond calculated the actual fabrication yield of each
control number using a proprietary method.  At verification, we found several
problems with this method.  However, because a description of this method is
not available on the public record of this case, we are unable to discuss it
here.  For a more detailed discussion this proprietary method and the problems
associated with its use, see the memorandum from Stan Bowen, et al., to Neal
Halper entitled "Calculation of Winbond's Fabrication Yields," dated September
5, 2000. 

As a result of our findings, we find that it is appropriate to adjust
Winbond's reported costs. Specifically, we have increased the reported costs
for all products (except those examined at verification (9)) by the percentage
difference between the weighted-average fabrication yield loss found for the
five models reviewed at verification and their reported yields.  As for using
adverse inferences, we agree with Winbond that it is not warranted in this
proceeding.  Section 776(b) of the Act states that the Department may use
adverse inferences when a party has failed to cooperate by not acting to the
best of its ability to comply with requests for information.  In contrast,
Winbond responded to all of the Department's requests for information in a
timely manner, and we verified the accuracy of this data.


With respect to Winbond's observation that the Department has accepted its
yield methodology in prior proceedings, we note that the Department may change
its position on a specific issue taken in prior proceedings if it provides an
explanation for the change (see, e.g.,  Final Results of Antidumping Duty
Administrative Review Elemental Sulphur From Canada; 64 FR 37737, 37740 (July
13, 1999) and Notice of Final Determination of Sales at Less Than Fair Value;
Certain Hot-Rolled Flat-Rolled Carbon-Quality Steel Products From Brazil; 64 FR
38756, 38787 (July 19, 1999).)  In this case, based upon our further
understanding of Winbond's methodology, we now find that Winbond's fabrication
yield calculation method creates distortive results, and as a consequence, we
have adjusted it as described above for purposes of the final results.


Comment 7: Variances


According to the petitioner, the Department found at verification that Winbond
excluded a number of variances from its cost calculations, including the
standard cost adjustment variances associated with beginning "WIP on hand,"
beginning finished goods, goods received, and goods replaced from the reported
costs.  Consequently, the petitioners claim that Winbond has understated its
reported costs, and as a result, the Department should increase Winbond's
reported costs by the amount of the understatement.


Winbond counters that it included all appropriate variances in its reported
costs.  Winbond claims that this issue arose from an observation in the cost
verification report that Winbond had excluded certain manufacturing variances
from the reported costs.  According to Winbond, this statement stemmed from a
misinterpretation of certain items identified on a submitted reconciliation
worksheet.  Winbond maintains that these items represent the differences
between the manufacturing costs actually incurred while products are being
manufactured and those normally recognized as part of the cost of manufacturing
in Winbond's accounts (e.g., although an R&D variance is incurred while
manufacturing products, it is considered to be an R&D expense and recorded as
such in Winbond's books).  Regarding the non-manufacturing variances identified
in the cost verification report and noted by the petitioner, Winbond states
that it excluded them to avoid double counting.  Specifically, Winbond notes
that it appropriately excluded the standard cost adjustment related to "WIP on
hand" (i.e., semi-finished goods) and finished goods because its cost
calculation methodology, which relies on the full standard unit cost of the
product, already captures these variances.  


Department's Position:


We agree with the respondent that record evidence shows that the company
appropriately included its manufacturing variances in its reported costs. 
Regarding the treatment of the standard cost variance (or adjustment)
associated with "WIP on hand," we agree with the petitioner that it should be
included in the calculation of COP and constructed value.  Consistent with our
general practice, Winbond excluded the standard cost variance associated with
finished goods inventories from the reported costs and included the adjustment
associated with WIP and raw materials.  (See, e.g., DRAMs from Taiwan, 64 FR at
56327.)  Winbond, however, excluded the variance associated with a portion of
its WIP inventory, described as "WIP on hand (semi-finished goods)," from the
reported costs.  We disagree that this cost variance should be excluded.  


For reporting purposes, Winbond differentiated its standard cost adjustment
associated with WIP into WIP and "WIP on hand" amounts.  According to Winbond,
the "WIP" category represented products undergoing the manufacturing process
but not yet fully completed, while the "WIP on hand" category accounts for
products that have completed a particular process and may either be sold or
further processed.  As noted earlier, Winbond included the "WIP" standard cost
adjustment in the reported costs and excluded the "WIP on hand" portion. 
Winbond reasons that a semi-finished good and a finished good are comparable
because both can be sold.  However, we disagree with Winbond's analogy because
record evidence does not support such a conclusion.


Specifically, we point out that this conclusion is not consistent with
Winbond's own accounting method used in the ordinary course of business.  In
its accounting and production control systems, Winbond considers "WIP on hand"
to be "WIP" rather than a component of finished goods.  For instance, Winbond's
audited financial statements show "WIP on hand" as being a component of WIP. 
Likewise, Winbond's production control system considers "WIP on hand" to be a
component of  WIP rather than finished goods.  At verification, we obtained and
reviewed sample production reports which traced Winbond's production flow. 
(See cost verification exhibit  6.)  These production reports identified
Winbond's "WIP on hand" as products pending completion (e.g., untested SRAMs,
SRAMs being further processed by subcontractors, SRAMs that require further
processing, etc.) rather than finished goods.  In addition, we note that
Winbond has based its reported costs on finished production quantities rather
than the cost of finished and "WIP on hand" quantities.  For instance, at
verification we reconciled the reported production quantities to Winbond's
finished goods ledgers.  (See cost verification exhibit 19.) We then reconciled
the amounts reported on the finished goods ledgers to the finished goods amount
reported on the company's audited financial statements.  As noted earlier, the
finished goods amount reported on the financial statement excludes "WIP on
hand."  


Despite the arbitrary classification, we must highlight that Winbond can and
does further process these semi-finished products at an additional cost.  Thus,
Winbond has not realized the full cost of manufacturing the product until it
has transferred the product to finished goods. Consequently, we conclude that
Winbond should not treat its "WIP on hand" standard cost adjustment differently
than its WIP standard cost adjustment.  For purposes of the final results, we
have included the standard cost adjustment variance associated with "WIP on
hand" in the calculation of COP and constructed value.
          
Comment 8: Foreign Exchange Losses Related to Cash Transactions
 
Winbond argues that the Department should exclude foreign exchange losses
related to cash transactions from the calculation of G&A expenses.  According
to Winbond, the Department's practice is to include only gains and losses
associated with manufacturing in G&A and to exclude gains and losses associated
with sales activities.  Winbond asserts that this practice is set forth in
Notice of Final Determination of Sales at Less Than Fair Value: Steel Wire Rod
From Trinidad & Tobago, 63 FR 9177, 9181 (Feb. 24, 1998) (Steel Wire Rod from
Trinidad & Tobago).  According to Winbond, in Steel Wire Rod from Trinidad &
Tobago the Department excluded foreign exchange gains related to accounts
receivable and cash accounts because they related to sales activities.  Winbond
concedes that in Final Determination of Sales at Less Than Fair Value:
Stainless Steel Round Wire from Canada, 64 FR, 17324, 17334 (Apr. 9, 1999)
(Round Wire from Canada), the Department found that exchange losses related to
cash should be included in the G&A ratio, but claims that this action appears
to have been an oversight. Winbond asserts that, in Round Wire from Canada, the
Department did not intend to modify its basic policy in this area because it
cited Steel Wire Rod from Trinidad & Tobago as support for its decision. 
Finally, Winbond asserts that any policy in which the Department includes
losses but not gains in G&A would be unfair and should not be followed.


According to the petitioner, the Department should include foreign exchange
losses related to cash in the calculation of Winbond's G&A expenses.  The
petitioner maintains that the Department's practice is to exclude foreign
exchange losses related to accounts receivable but to include those related to
cash.  As support for this assertion, the petitioner cites Round Wire from
Canada which states that only the foreign-exchange gains and losses that relate
to maintaining accounts payable and cash accounts were included in COP and
constructed value.  According to the petitioner, Winbond's reliance on Steel
Wire Rod from Trinidad & Tobago is misplaced because in that case the
Department found that the exchange losses in question related to sales
transactions.  The petitioner maintains that, because Winbond has made no
showing that its losses on cash were related in any way to sales transactions,
the Department should continue to include them in G&A expenses.
       
Department's Position:
    
It is the Department's practice to exclude from general expenses only those
foreign exchange gains and losses which are associated with sales transactions.
See Round Wire from Canada, 64 FR at 17334.  For example, the Department
excludes exchange gains and losses associated with accounts receivable because
accounts receivable balances are generated when sales are made. Foreign
exchange gains and losses arising from cash transactions, on the other hand,
may be associated with non-sales transactions. (10)   We disagree with Winbond
that this practice is inconsistent with the policy set forth in Steel Wire Rod
from Trinidad & Tobago.  In that case, we merely stated that it is our normal
practice to exclude exchange gains and losses from sales transactions from the
calculation of COP and constructed value and to include exchange gains and
losses arising from purchase transactions.  See Steel Wire Rod from Trinidad &
Tobago, 63 FR at 9182.


At verification, Winbond did not demonstrate that the exchange losses in
question were linked to the company's sales, nor is there any other information
on the record which establishes that this is the case.  Consequently,
consistent with our established practice, we have continued to include these
losses in the calculation of the COP and constructed value for purposes of the
final results.


Comment 9: Research and Development Costs             


In the preliminary results, the Department allocated total semiconductor R&D
expenses over total semiconductor cost of sales.  Winbond argues that the
Department should recalculate its R&D expenses on an SRAM-specific basis
because the company tracks R&D expenses at this level in its normal accounting
system.  Specifically, Winbond argues that the Department should base the
company's R&D expenses on the costs reported for SRAM-related R&D projects in
1998.


Winbond acknowledges that the Department has allocated semiconductor R&D over
total semiconductor cost of sales in prior semiconductor cases, including the
original investigation in this proceeding.  According to Winbond, the
Department reasons that "cross fertilization" exists in the semiconductor
industry, and that R&D costs incurred for any semiconductor product, no matter
how far removed from the subject merchandise, benefit the design, development,
and production of the subject merchandise.  However, Winbond argues that the
Department should reconsider this approach because it is not reasonable. 
Specifically, Winbond asserts that the R&D costs incurred on logic products,
which represent a substantial portion of the company's business, do not impact
the development of its SRAM products.  Winbond states that it demonstrated at
verification that the company's logic R&D activities have an entirely different
design focus and are more complex from an engineering perspective than are its
SRAM R&D activities.  For example, Winbond states that its logic R&D is geared
toward designing an application-specific device to a customer's specific design
requirements, whereas SRAM R&D is focused on increasing data processing
capacity and efficiency.  Winbond asserts that SRAM process technology is
generally more advanced than logic process development, and thus by definition
could not benefit from logic R&D.  Winbond notes that there are no specific
instances in the record of this proceeding of cross-fertilization of R&D at
Winbond. 


Winbond finally agues that use of SRAM-specific R&D costs would be consistent
with the Department's preference for using product-specific costs to the extent
possible.  Winbond asserts that this consideration is particularly valid in
this case, given that R&D costs are recorded in Winbond's accounting system at
a product-specific level in the normal course of business.


The petitioner asserts that none of the arguments presented above are new -
Winbond raised each of these arguments in the LTFV investigation and the
Department considered them then. According to the petitioner, because there are
no new facts in this segment of the proceeding, the Department should continue
to follow its long-standing practice in semiconductor cases of allocating R&D
for all semiconductor products over the cost of sales of such products.


Department's Position:


We agree with the petitioner and have continued to find that there is cross-
fertilization of scientific ideas between the R&D activities of semiconductor
products.  As we noted in theNotice of Final Determination of Sales at Less
Than Fair Value: Static Random Access Memory Semiconductors From Taiwan, 63 FR
8909, 8924 (Feb. 23, 1998) (SRAMs Final Determination), processing advancements
for one semiconductor product can benefit other types of semiconductor products
(including logic and memory).  Furthermore, design improvements, although
undertaken for a specific product, can and often do, become incorporated into
the design of other semiconductors, whether they are logic or memory devices.
We find that it is appropriate to allocate the cost of all semiconductor R&D to
all semiconductor products, given that scientific ideas developed in one
semiconductor area can be and have been utilized in the development of other
semiconductor products. Therefore, for purposes of the final results, we have
calculated R&D for SRAMs using the ratio of total semiconductor R&D to total
semiconductor cost of sales for the most recently completed fiscal period for
which the respondents have finalized financial statements.


Due to the forward-looking nature of R&D activities, the Department cannot
identify every instance where SRAM R&D may influence logic products or where
logic R&D may influence SRAM products.  However, we note that in the LTFV
investigation, the Department's own expert identified areas where R&D from one
type of semiconductor product has influenced another semiconductor product. 
Dr. Murzy Jhabvala, a semiconductor device engineer at NASA with twenty-four
years of experience, was invited by the Department to express his views
regarding cross-fertilization of R&D efforts in the semiconductor industry.  He
stated that "it is reasonable and realistic to contend that R&D from one area
(e.g., bipolar) applies and benefits R&D efforts in another area (e.g., MOS
memory).''  Dr. Jhabvala went on to state that "SRAMs represent along with
DRAMs the culmination of semiconductor research and development.  Both families
of devices have benefitted from the advances in photolithographic techniques to
print the fine geometries (the state-of-the-art steppers) required for the high
density of transistors..."  Clearly, three distinct areas of semiconductor
technology are converging to benefit the SRAM device performance.   See SRAMs
Final Determination, 63 FR at 8924.


There are other instances where previous technology and the efforts expended
to develop that technology occurs in the SRAM technology.  Some examples of
these are the use of thin film transistors in SRAMs, advanced metal
interconnect systems, anisotropic etching and filling techniques for trenching
and planarization and implant technology for retrograde wells.  In the LTFV
investigation, the Department also identified through published magazine
articles examples of cross-fertilization in the semiconductor industry.  See,
e.g., "A 250-MHz Skewed-Clock Pipelined Data Buffer,'' Institute of Electrical
and Electronics Engineers Journal of Solid State Circuits, March 1996; and "A 1-
Mb 2 Tr/b Nonvolatile CAM Based on Flash Memory Technologies,'' Institute of
Electrical and Electronics Engineers Journal of Solid State Circuits, November
1996.  We also noted numerous published articles in the Institute of Electrical
and Electronics Engineers Journal of Solid State Circuits which described how
significant advancements in the advanced semiconductor integrated circuit
(ASIC)/logic product area have had important ramifications for chip design in
the memory areas.  The articles described how multilayer metal design
development categorized as logic/ASIC R&D will permit companies to build chips
that are smaller, faster and more power-efficient.  The articles concluded that
the research will be used in the future to improve microprocessors, memory and
mixed-signal devices.  As an example, one article entitled "The Challenges of
Embedded DRAM in ASICs: A Manufacturing Economics Point of View,'' Dataquest
Interactive, August 25, 1997, discussed the technical challenges of embedding
memory into ASICs, which illustrated the overlap in design and process
technology between logic and memory circuits.  This article noted on page two
that "[b]oth the fast SRAM and the 'pseudo-DRAM' structures are actually
subsets of the process flow for advanced logic, so designing and constructing
SLI ASICs are a natural extension and do not really add much to the per-wafer
cost of the process.''  The articles are attached as exhibits to a letter
submitted by the petitioner on September 17, 1999, as well as in the memorandum
to The File from Irina Itkin, entitled "Articles from Semiconductor Research
Periodicals," dated September 5, 2000.


We also disagree with Winbond's assertion that the methodology employed by the
Department should be based on respondents' normal accounting records.  While we
do not disagree that each R&D project is accounted for separately in Winbond's
books and records, we note that the existence of separate accounting records
does not necessarily preclude the phenomenon of cross-fertilization of
scientific ideas.  Since accounting records do not address the critical issue
of whether ideas from research in one area benefit another area, we do not find
this argument persuasive.  


We also find unpersuasive Winbond's arguments that logic R&D is more complex
than memory R&D and that logic R&D is unique to an application.  The record
shows that the primary focus for SRAM R&D is reducing die size and increasing
speed, which will benefit from the metal multilayer design R&D being conducted
in connection with logic/ASIC products. Moreover, the issue is not whether
application-specific design R&D for logic products can be used for SRAMs, but
rather whether what is learned from logic/ASIC product R&D can be used to
improve SRAM performance.


Based on the foregoing, for purposes of the final results, we have continued
to calculate R&D for SRAMs using the ratio of total semiconductor R&D to total
semiconductor cost of sales for the fiscal year beginning in 1998, consistent
with our treatment of R&D costs in the preliminary results.


Comment 10:  Products Produced But Not Sold During the Review Period


According to the petitioner, the reconciliation performed at the cost
verification at Winbond was suspect and thus should be rejected in part. 
Specifically, the petitioner asserts that Winbond's reconciliation worksheet
reflected a significant value of products which were produced during the POR
but not sold during this period; however, the petitioner states that these
products do not appear to be reflected in the inventory values shown on the
company's audited financial statements.  The petitioner claims that Winbond has
not demonstrated that the products that it identified on this worksheet as
"produced but not sold" were ever actually sold.  Further, the petitioner
argues that, if Winbond is not able to demonstrate this, the Department should
allocate the cost of producing these SRAMs over the cost of the SRAMs that were
sold during the POR.


According to Winbond, the petitioner's analysis of the product costs in
question is erroneous. Winbond notes that the Department thoroughly verified
the reported costs, including the reconciliation of production and sales
quantities for the entire year of 1998, which accounts for two-thirds of the
POR.  Winbond asserts that the petitioner has ignored this and instead points
to the data for 1997, for which the Department did not request support at
verification.  Winbond argues that it would be unreasonable for the Department
to make an adverse inference in this case because Winbond supposedly failed to
prove something that the Department never requested.  Winbond asserts that, had
the Department requested the information at verification, it would have
provided it then.


Nonetheless, Winbond maintains that the differences in the figures are easily
explained.  First, Winbond notes that not all products produced by the company
are sold to outside parties because some may be consumed internally in the
manufacture of other products.  Moreover, Winbond notes that the cost files did
not contain data for all products manufactured in 1997, but rather only those
products sold during the last quarter of that year.  Thus, Winbond asserts that
a difference of the magnitude noted by the petitioner is not surprising. 
Accordingly, Winbond asserts that the Department should accept its data as
reported.


Department's Position:


At verification, Winbond provided worksheets which reconciled the costs
recorded in its accounting system during the POR to the costs reported in its
COP and constructed value databases.  We verified the accuracy of these
worksheets by selecting a number of costs shown on them and tracing these costs
to the financial statements and to Winbond's response.  We noted no
discrepancies during our review of these documents.


Although we did not specifically examine the costs referenced by the
petitioner in its case brief, we find that this fact does not call the
verification process into question in any way.  In order to conduct its
verifications in the time allotted, the Department is frequently required to
select the documents that it reviews on a sample basis.  In situations where:
1) the selected data reported by the respondent ties to its accounting system
without discrepancy; and 2) the respondent cooperates fully during the
verification process by complying with each request for information, we
generally do not find it necessary to obtain supporting documentation for every
reconciling item.  Thus, because we were satisfied at verification with the
reconciliations performed by Winbond, we have accepted Winbond's data without
adjustment (except where otherwise noted elsewhere).


Comment 11: Bonuses


In the preliminary results, the Department adjusted Winbond's constructed
value to include a stock bonus for the first quarter of 1999.  Because Winbond
had not declared a stock bonus in this quarter, we based the amount of this
bonus on the bonus rate paid in 1998.  According to Winbond, this calculation
was not appropriate because the company did not declare, or pay, stock bonuses
in 1999.


Department's Position:


At verification, we found no evidence that Winbond paid stock bonuses to its
employees during 1999.  Consequently, because Winbond did not incur any costs
related to stock bonuses during this period, we revised our calculations to
remove these costs from constructed value for purposes of the final results. 
For further discussion, see Comment 12, below.


Comment 12: Clerical Errors in the Preliminary Results


According to the petitioner, the Department made two clerical errors in the
preliminary margin calculations for Winbond.  Specifically, the petitioner
claims that the Department computed stock bonuses for the POR, but did not
include these costs in the COP.  Similarly, the petitioner asserts that the
Department calculated U.S. warehousing costs, but did not deduct these amounts
from CEP.  The petitioner requests that the Department correct these errors for
purposes of the final results.


Winbond argues that the Department should make no adjustment to the COP for
bonus expenses for 1999, because it declared no stock bonuses in that year. 
See Comment 11.


Department's Position:


We agree that we inadvertently failed to deduct U.S. warehousing expenses from
CEP.  We also agree that we failed to include include stock bonuses for 1997
and 1998 in the COP. Consequently, we have revised our calculations for
purposes of the final results to correct these errors.  Regarding 1999 bonuses,
however, we agree with Winbond that these expenses should not be included in
COP or constructed value.  Therefore, we have made no corrections to our
calculations for purposes of the final results for 1999 stock bonus expenses. 
For further discussion, see Comment 11, above.


  
Comment 13: Constructed Export Price Offset

The Department granted a CEP offset to GSI Technology and Winbond for purposes
of the preliminary results.   According to the petitioner, the Department
should deny this offset for purposes of the final results because the
Department's analysis is contrary to law.
 
The petitioner notes that, in two recent decisions, the Court of International
Trade (CIT) has held that the Department must perform its level of trade (LOT)
analysis based on unadjusted starting prices for both U.S. and home market
sales.  See Borden, Inc. v. United States, 4 F. Supp. 2d 1221 (1998) (Borden). 
See also Micron Technology, Inc. v. United States, 40 F. Supp. 2d 481, 485-486
(1999) (Micron).  The petitioner contends that the Department's analysis is
contrary to these rulings because the Department conducted its analysis based
on the level of the constructed sale from the exporter to the importer (i.e.,
after adjustment for U.S. indirect selling expenses). According to the
petitioner, when the Department conducts a corrected LOT analysis, it will find
that the comparison market sales made by GSI Technology and Winbond were not
made at a more advanced LOT than their sales in the United States, and,
therefore, there is no basis for granting a CEP offset to either respondent.


Winbond disagrees, stating that the Department's consistent policy is to
analyze the level of trade of U.S. sales based on the level of the constructed
sales from the exporter to the affiliated importer.  Winbond asserts that there
is nothing in the law or the facts of this administrative review to suggest
that the Department should reexamine this practice.  Finally, Winbond notes
that the Department has stated that it disagrees with Borden and Micron and
that it has not followed those decisions in subsequent cases.  For example,
Winbond cites Industrial Nitrocellulose From the United Kingdom; Notice of
Final Results of Antidumping Duty Administrative Review, 65 FR 6148, 6150 (Feb.
8, 2000), in which the Department stated that it will "continue to analyze the
level of trade based on adjusted CEP prices rather than the starting CEP
prices."  Thus, Winbond asserts that there is no reason to revisit the
Department's preliminary finding that Winbond is entitled to a CEP offset. 


GSI Technology did not comment on this issue.


Department's Position:


We recognize that the CIT has held that the Department's practice of
determining LOTs for CEP transactions after CEP deductions is an impermissible
interpretation of section 772(d) of the Act. See Borden, 4 F. Supp. 2d at 1241-
1242. (11)  The Department believes, however, that its practice is in full
compliance with the statute.  On June 4, 1999, the CIT entered a final
judgement inBorden on the LOT issue.  See Borden, Inc., v. United States, Court
No. 96-08-01970, Slip Op. 99-50 (CIT June 4, 1999).  The government has filed
an appeal of Borden which is pending before the U.S. Court of Appeals for the
Federal Circuit.  Consequently, the Department has continued to follow its
normal practice of adjusting CEP under section 772(d) of the Act prior to
starting a LOT analysis, as articulated in the Department's regulations at
section 351.412.


The Department has consistently stated that the Act and the SAA support
analyzing the level of trade of CEP sales at the constructed level, after
expenses associated with economic activities occurring in the United States
have been deducted pursuant to section 772(d) of the Act.  In the preamble to
our proposed regulations, we stated:


 With respect to the identification of levels of trade, some commentators
argued that, consistent with past practice, the Department should base level of
trade on the starting price for both export price EP and CEP sales...  The
Department believes that this proposal is not supported by the SAA.  If the
starting price is used for all U.S. sales, the Department's ability to make
meaningful comparisons at the same level of trade (or appropriate adjustments
for differences in levels of trade) would be severely undermined in cases
involving CEP sales.  As noted by other commentators, using the starting price
to determine the level of trade of both types of U.S. sales would result in a
finding of different levels of trade for an EP sale and a CEP sale adjusted to
a price that reflected the same selling functions.  Accordingly, the
regulations specify that the level of trade analyzed for EP sales is that of
the starting price, and for CEP sales it is the constructed level of trade of
the price after the deduction of U.S. selling expenses and profit.


See Antidumping Duties; Countervailing Duties; Notice of Proposed Rule Making
and Request for Public Comments, 61 FR 7308, 7347 (Feb. 27, 1996).


Consistent with the above position, in those cases where a level of trade
comparison is warranted and possible, the Department evaluates the level of
trade for CEP sales based on the price after adjustments are made under section
772(d) of the Act.  See, e.g., Large Newspaper Printing Presses and Components
Thereof, Whether Assembled of Unassembled, From Japan: Notice of Final
Determination of Sales at Less Than Fair Value, 61 FR 38139, 38143 (July 23,
1996).  We note that, in every case decided under the revised antidumping
statute, we have consistently adhered to this interpretation of the SAA and of
the Act.  See, e.g., Extruded Rubber Thread From Malaysia; Final Results of
Antidumping Duty Administrative Review, 65 FR 6140, 6141 (Feb. 8, 2000); Notice
of Final Determination of Sales at Less Than Fair Value: Dynamic Random Access
Memory Semiconductors of One Megabit and Above ("DRAMs") from Taiwan, 64 FR
56308, 56313 (Oct. 19, 1999); Antifriction Bearings (Other Than Tapered Roller
Bearings) and Parts Thereof from France, Germany, Italy, Japan, Singapore, and
the United Kingdom; Final Results of Antidumping Duty Administrative Reviews,
62 FR 2081, 2106 (Jan. 15, 1997); and Aramid Fiber Formed of Poly Para-
Phenylene Terephthalamide from the Netherlands; Final Results of Antidumping
Administrative Review, 61 FR 51406, 51408 (Oct. 2, 1996). 


In this case, in accordance with the above precedent, our instructions in the
questionnaire issued to GSI Technology and Winbond stated that the constructed
level of trade should be used.  We verified the differences in selling
functions in the foreign and U.S. markets, after all selling expenses incurred
in the United States were deducted from the CEP.  Specifically, we found that
all of the selling functions related to all of GSI Technology's U.S. and third
country market sales, as well as the majority of Winbond's selling functions
related to U.S. sales, occurred in the United States.  See the GSI USA sales
verification report, dated April 7, 2000, at page 2.  Seealso the Winbond USA
sales verification report, dated April 6, 2000, at page 21.


As stated in the SAA, a constructed export price offset adjustment will be
made only where: 1) the data available do not form an appropriate basis for
determining an LOT adjustment under section 773(a)(7)(A)(ii) of the Act; and 2)
normal value is established at a level of trade more remote from the factory
than the level of trade of the CEP.  Because we find that only one level of
trade existed in the foreign markets during the POR, we are unable to make a
LOT adjustment under section 773(a)(7)(A)(ii) of the Act.  See Preliminary
Results, 65 FR at 26579.  Moreover, because normal value is based on a price
which reflects full selling activities, while the CEP after adjustment is based
on a price which reflects few (for Winbond) or none (for GSI Technology), we
find that the normal value is at a level of trade more remote from the factory
than the level of trade of the CEP.  Accordingly, we find that the Department's
decision to grant a CEP offset to these respondents is consistent with the
statute and that the Department's practice is supported by substantial evidence
on the record.  Consequently, consistent with our preliminary results, we have
continued to analyze the level of trade based on adjusted CEP prices, rather
than the starting CEP prices, and we have continued to grant a CEP offset to
GSI Technology and Winbond for purposes of the final results.



Recommendation


Based on our analysis of the comments received, we recommend adopting all of
the above positions.  If these recommendations are accepted, we will publish
the final results of review and the final weighted-average dumping margins for
the reviewed firms in the Federal Register.




Agree____     Disagree____


                            
                                          
Troy H. Cribb
Acting Assistant Secretary for
  Import Administration
 
 _________________________________________________________________________    
                                         
Footnotes:          
1.  G-Link disagrees with a statement in the cost report which indicates that
the company could have calculated SRAM-specific variances.  According to G-
Link, each of the exhibits cited by the Department as support for this
statement does not contain sufficient information to compute product-specific
variances.
2.   According to G-Link, this adjustment related to a reduction in the value
of finished goods inventory, due to the fact that the company's standard costs
were higher in prior quarters.
3.  In that case, the Department accepted the respondent's reported variances
that were either product-group- or company- wide.  The Department noted that
the respondent's cost accounting system normally applied company-wide
variances.  However, for submission purposes, the respondent converted a
company-wide variance into distinct product-group variances because its cost
accounting system maintained the necessary information to make such a
calculation.
4.  In similar instances, the courts have upheld our use of partial facts
available.  SeeRautaruukki Oy. V. United States, Slip Op. 98-112 (Aug. 4,
1998), where the court agreed that the Department's use of partial facts
available was reasonable because the respondent was cooperative and the
adjustment was supported by record evidence.  In that case, as partial facts
available, the Department used the highest reported "cost extra" for certain
products.  
5.  We agree with G-Link that it is appropriate to exclude its standard cost
variance associated with finished goods inventory from the reported costs. 
However, we note that our general practice is to include the write-down
associated with raw material and WIP in the calculation of COP and constructed
value.  (See, e.g., Notice of Final Results of Antidumping Duty Administrative
Review and Determination Not to Revoke the Antidumping Duty Order: Brass Sheet
and Strip From the Netherlands, 65 FR 742-752 (Jan. 6, 2000) and Final
Determination of Sales at Less Than Fair Value: Dynamic Random Access Memory
Semiconductors of One Megabit and Above (DRAMs) From Taiwan, 64 FR 56308, 56326
(Oct. 19, 1999) (DRAMs from Taiwan).)  Therefore, we have adjusted G-Link's
reported costs to include a portion of the standard cost adjustment associated
with raw materials and WIP.  
6.   For example, of the 105 sales reported in the third country database,
approximately 10 percent of the sales were made in the same or smaller
quantities as two of the three sales in question and over a third of the sales
were made in the same or smaller quantities as the remaining sale.
7.  The petitioner also notes that Winbond's U.S. sales volume during this POR
represented only one percent its home market sales volume. 
8.   Due to the fact that Winbond claimed business proprietary treatment for
the nature of this data, we are unable to discuss it in this memorandum.  See
pages 1 and 2 of the petitioner's case brief and pages 4 and 5 of Winbond's
rebuttal brief for the details of this calculation.  In addition, we have
addressed the parties' arguments in a separate memorandum to the file, which
will be placed on the record and served upon parties with access to such
information under administrative protective order.  See memorandum from the
Stan Bowen, et al., to Neal Halper entitled "Calculation of Winbond's
Fabrication Yields," dated September 5, 2000.
9.  For these products, we adjusted the costs according to our findings at
verification.
10.   For example, companies may incur and service debt obligations in U.S.
dollars or they may purchase inputs using overseas foreign currency-denominated
bank accounts.  
11.   We also note that the CIT declined to follow the CEP offset rationale
set forth inBorden in a subsequent case.  See NTN Bearing Corp. v. U.S., 2000
Ct. Intl. Trade LEXIS 67, Slip Op. 2000-64 (CIT June 5, 2000).